Marcel Finan S Exam 1P Manual

A Probability Course for the Actuaries A Preparation for Exam P/1 Marcel B.
Finan Arkansas Tech University c All Rights Reserved Preliminary Draft
2 In memory of my parents August 1, 2008 January 7, 2009
Preface The present manuscript is designed mainly to help students prepare for the Proba bility Exam (Exam P/1), the rst actuarial examination administered by the Society of Actuaries. This examination tests a students knowledge of the fundamental pro bability tools for quantitatively assessing risk. A thorough command of calculus is assumed. More information about the exam can be found on the webpage of the Society of Actauries www.soa.org. Problems taken from samples of the Exam P/1 pr ovided by the Casual Society of Actuaries will be indicated by the symbol . The ow of topics in the book follows very closely that of Rosss A First Course in Proba biltiy. This manuscript can be used for personal use or class use, but not for c ommercial purposes. If you nd any errors, I would appreciate hearing from you: m na n@atu.edu This manuscript is also suitable for a one semester course in an under graduate course in probability theory. Answer keys to text problems are found at the end of the book. This project has been partially supported by a research gr ant from Arkansas Tech University. Marcel B. Finan Russellville, AR May 2007 3
4 PREFACE
Contents Preface 3 Basic Operations on Sets 9 1 Basic De nitions . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2 Set Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Counting and Combinatorics 31 3 The Fundamental Principle of Cou nting . . . . . . . . . . . . . . 31 4 Permutations and Combinations . . . . . . . . . . . . . . . . . . . 37 5 Permutations and Combinations with Indistinguish able Objects . 47 Probability: De nitions and Properties 6 Basic De nitions and Axio ms of Probability . . . . . . . . . . . . 7 Properties of Probability . . . . . . . . . . . . . . . . . . . . . . . 8 Probability and Counting Techniques . . . . . . . . . . . . . . . . Conditional Probability and Independence 9 Conditional Probabilities . . . . . . . . . . . 10 Posterior Probabilities: Bayes Formula . . 11 Independent Events . . . . . . . . . . . . . 12 Odds and Conditional Probab ility . . . . . 57 57 65 73 79 79 87 98 107 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Discrete Random Variables 111 13 Random Variables . . . . . . . . . . . . . . . . . . . . . . . . . . 111 14 Probability Mass Function and Cumulative Distributi on Function117 15 Expected Value of a Discrete Random Variable . . . . . . . . . . 125 16 Expected Value of a Function of a Discrete Random Variable . . 132 17 Variance and Standard Deviation . . . . . . . . . . . . . . . . . 139 5
6 18 Binomial and Multinomial Random Variables . . . . 19 Poisson Random Variabl e . . . . . . . . . . . . . . . 20 Other Discrete Random Variables . . . . . . . . . . 20.1 Geometric Random Variable . . . . . . . . . 20.2 Negative Binomial R andom Variable . . . . . 20.3 Hypergeometric Random Variable . . . . . . 21 Prop erties of the Cumulative Distribution Function . . . . . . . . . . . . . . CONTENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 45 159 169 169 176 183 189 Continuous Random Variables 22 Distribution Functions . . . . . . . . . . . . . . . . . 23 Expectation, Variance and Standard Deviation . . . . 24 The Uniform D istribution Function . . . . . . . . . . 25 Normal Random Variables . . . . . . . . . . . . . . . 26 Exponential Random Variables . . . . . . . . . . . . . 27 G amma and Beta Distributions . . . . . . . . . . . . 28 The Distribution of a Fun ction of a Random Variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203 . 203 . 216 . 234 . 239 . 254 . 264 . 276 Joint Distributions 285 29 Jointly Distributed Random Variables . . . . . . . . . . . . . . . 285 30 Independent Random Variables . . . . . . . . . . . . . . . . . . 299 31 Sum of Two Independent Random Variables . . . . . . . . . . . 311 3 1.1 Discrete Case . . . . . . . . . . . . . . . . . . . . . . . . 311 31.2 Conti nuous Case . . . . . . . . . . . . . . . . . . . . . . . 316 32 Conditional Dist ributions: Discrete Case . . . . . . . . . . . . . 325 33 Conditional Distributi ons: Continuous Case . . . . . . . . . . . . 332 34 Joint Probability Distributi ons of Functions of Random Variables 341 Properties of Expectation 35 Expected V alue of a Function of Two Random Variables . 36 Covariance, Variance of Sums, an d Correlations . . . . . 37 Conditional Expectation . . . . . . . . . . . . . . . . . . 38 Moment Generating Functions . . . . . . . . . . . . . . . Limit Theor ems 39 The Law of Large Numbers . . . . . . . . 39.1 The Weak Law of Large Numbe rs 39.2 The Strong Law of Large Numbers 40 The Central Limit Theorem . . . . . . . 349 . 349 . 359 . 372 . 383 397 . 397 . 397 . 403 . 414 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . .
CONTENTS 7 41 More Useful Probabilistic Inequalities . . . . . . . . . . . . . . . 424 Appe ndix 431 42 Improper Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . 431 43 Double Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . 438 44 Double Integrals in Polar Coordinates . . . . . . . . . . . . . . . 451 Answer Keys BIBLIOGRAPHY 457 508
8 CONTENTS
Basic Operations on Sets The axiomatic approach to probability is developed using the foundation of set t heory, and a quick review of the theory is in order. If you are familiar with se t builder notation, Venn diagrams, and the basic operations on sets, (unions, in tersections, and complements), then you have a good start on what we will need r ight away from set theory. Set is the most basic term in mathematics. Some synon yms of a set are class or collection. In this chapter we introduce the concept o f a set and its various operations and then study the properties of these operat ions. Throughout this book, we assume that the reader is familiar with the follo wing number systems: The set of all positive integers N = {1, 2, 3, }. The set o f all integers Z = { , 3, 2, 1, 0, 1, 2, 3, }. The set of all rational numb { : a, b Z with b = 0}. b The set R of all real numbers. 1 Basic De nitions We de ne a set A as a collection of well de ned objects (called elements or members of A) such that for any given object x either one (but not both) 9
10 of the following holds: x belongs to A and we write x A. BASIC OPERATIONS ON SETS x does not belong to A, and in this case we write x A. Example 1.1 Which of the following is a well de ned set. (a) The collection of good books. (b) The collecti on of left handed individuals in Russellville. Solution. (a) The collection of g ood books is not a well de ned set since the answer to the question Is My Life a go od book? may be subject to dispute. (b) This collection is a well de ned set since a person is either left handed or right handed. Of course, we are ignoring those few who can use both hands There are two di erent ways to represent a set. The rst one is to list, without repetition, the elements of the set. For example, if A is the solution set to the equation x2 4 = 0 then A = {2, 2}. The other way to re present a set is to describe a property that characterizes the elements of the s et. This is known as the set builder representation of a set. For example, the s et A above can be written as A = {x x is an integer satisfying x2 4 = 0}. We de ne the empty set, denoted by , to be the set with no elements. A set which is not e mpty is called a nonempty set. Example 1.2 List the elements of the following se ts. (a) {x x is a real number such that x2 = 1}. (b) {x x is an integer such tha t x2 3 = 0}. Solution. (a) {1, 1}. (b) Since the only solutions to the given equa tion are 3 and 3 and both are not integers, the set in question is the empty set Example 1.3 Use a property to give a description of each of the following sets. (a) {a, e, i, o, u}. (b) {1, 3, 5, 7, 9}.
1 BASIC DEFINITIONS Solution. (a) {x x is a vowel}. (b) {n N n is odd and less t han 10 } 11 The rst arithmetic operation involving sets that we consider is the equality of t wo sets. Two sets A and B are said to be equal if and only if they contain the s ame elements. We write A = B. For non equal sets we write A = B. In this case, t he two sets do not contain the same elements. Example 1.4 Determine whether each of the following pairs of sets are equal. (a) {1, 3, 5} and {5, 3, 1}. (b) {{1} } and {1, {1}}. Solution. (a) Since the order of listing elements in a set is ir relevant, {1, 3, 5} = {5, 3, 1}. (b) Since one of the sets has exactly one membe r and the other has two, {{1}} = {1, {1}} In set theory, the number of elements in a set has a special name. It is called the cardinality of the set. We write n (A) to denote the cardinality of the set A. If A has a nite cardinality we say th at A is a nite set. Otherwise, it is called in nite. For in nite set, we write n(A) = . For example, n(N) = . Can two in nite sets have the same cardinality? The answer is yes. If A and B are two sets ( nite or in nite) and there is a bijection from A t o B (i.e. a one to one and onto function) then the two sets are said to have the same cardinality, i.e. n(A) = n(B). Example 1.5 What is the cardinality of each of the following sets? (a) . (b) {}. (c) {a, {a}, {a, {a}}}. Solution. (a) n() = 0 .
12 BASIC OPERATIONS ON SETS (b) This is a set consisting of one element . Thus, n({}) = 1. (c) n({a, {a}, {a, {a}}}) = 3 Now, one compares numbers using inequalities. The corresponding notio n for sets is the concept of a subset: Let A and B be two sets. We say that A is a subset of B, denoted by A B, if and only if every element of A is also an ele ment of B. If there exists an element of A which is not in B then we write A B. For any set A we have A A. That is, every set has at least two subsets. Also, ke ep in mind that the empty set is a subset of any set. Example 1.6 Suppose that A = {2, 4, 6}, B = {2, 6}, and C = {4, 6}. Determine which of these sets are subs ets of which other of these sets. Solution. B A and C A If sets A and B are repr esented as regions in the plane, relationships between A and B can be represente d by pictures, called Venn diagrams. Example 1.7 Represent A B C using Venn diag ram. Solution. The Venn diagram is given in Figure 1.1 Figure 1.1 Let A and B be two sets. We say that A is a proper subset of B, denot ed by A B, if A B and A = B. Thus, to show that A is a proper subset of B we mus t show that every element of A is an element of B and there is an element of B w hich is not in A.
1 BASIC DEFINITIONS Example 1.8 Order the sets of numbers: Z, R, Q, N using Solu tion. NZQR Example 1.9 Determine whether each of the following statements is true o r false. (a) x {x} (b) {x} {x} (c) {x} {x} (d) {x} {{x}} (e) {x} (f) {x} 13 Solution. (a) True (b) True (c) False since {x} is a set consisting of a single element x and so {x} is not a member of this set (d) True (e) True (f) False sin ce {x} does not have as a listed member Now, the collection of all subsets of a set A is of importance. We denote this set by P(A) and we call it the power set of A. Example 1.10 Find the power set of A = {a, b, c}. Solution. P(A) = {, {a}, {b}, {c}, {a, b}, {a, c}, {b, c}, {a, b, c}} We conclude this section, by introd ucing the concept of mathematical induction: We want to prove that some statemen t P (n) is true for any nonnegative integer n n0 . The steps of mathematical ind uction are as follows: (i) (Basis of induction) Show that P (n0 ) is true. (ii) (Induction hypothesis) Assume P (n0 ), P (n0 + 1), , P (n) are true. (iii) (Indu ction step) Show that P (n + 1) is true. Example 1.11 (a) Use induction to show that if n(A) = n then n(P(A)) = 2n , where n 0 and n N. (b) If P(A) has 256 elem ents, how many elements are there in A?
14 BASIC OPERATIONS ON SETS Solution. (a) We apply induction to prove the claim. If n = 0 then A = and in th is case P(A) = {}. Thus, n(P(A)) = 1 = 20 . As induction hypothesis, suppose that if n(A) = n then n(P(A)) = 2n . Let B = {a1 , a2 , , an , an+1 }. Then P(B) con sists of all subsets of {a1 , a2 , , an } together with all subsets of {a1 , a2 , , an } with the element an+1 added to them. Hence, n(P(B)) = 2n + 2n = 2 2n = 2n+1 . (b) Since n(P(A)) = 256 = 28 , by (a) we have n(A) = 8 Example 1.12 Use i nduction to show that n i=1 (2i 1) = n2 , n N. Solution. If n = 1 we have 12 = 2(1)1 = 1 (2i1). Suppose that the result is true i =1 for up to n. We will show that it is true for n + 1. Indeed, n+1 (2i 1) = i=1 n (2i 1) + 2(n + 1) 1 = n2 + 2n + 2 1 = (n + 1)2 i=1
1 BASIC DEFINITIONS 15 Practice Problems Problem 1.1 Consider the experiment of rolling a die. List the elements of the s et A = {x : x shows a face with prime number}. Recall that a prime number is a n umber with only two divisors: 1 and the number itself. Problem 1.2 Consider the random experiment of tossing a coin three times. (a) Let S be the collection of all outcomes of this experiment. List the elements of S. Use H for head and T fo r tail. (b) Let E be the subset of S with more than one tail. List the elements of E. (c) Suppose F = {T HH, HT H, HHT, HHH}. Write F in set builder notation. P roblem 1.3 Consider the experiment of tossing a coin three times. Let E be the c ollection of outcomes with at least one head and F the collection of outcomes of more than one head. Compare the two sets E and F. Problem 1.4 A hand of 5 cards is dealt from a deck. Let E be the event that the hand contains 5 aces. List th e elements of E. Problem 1.5 Prove the following properties: (a) Re exive Property : A A. (b) Antisymmetric Property: If A B and B A then A = B. (c) Transitive Pro perty: If A B and B C then A C. Problem 1.6 Prove by using mathematical inductio n that 1 + 2 + 3 + + n = n(n + 1) , n N. 2 Problem 1.7 Prove by using mathematical induction that 12 + 22 + 32 + + n2 = n(n + 1)(2n + 1) , n N. 6
16 BASIC OPERATIONS ON SETS Problem 1.8 Use induction to show that (1 + x)n 1 + nx for all n N, where x > 1. Problem 1.9 A caterer prepared 60 beef tacos for a birthday party. Among these t acos, he made 45 with tomatoes, 30 with both tomatoes and onions, and 5 with nei ther tomatoes nor onions. Using a Venn diagram, how many tacos did he make with (a) tomatoes or onions? (b) onions? (c) onions but not tomatoes? Problem 1.10 A dormitory of college freshmen has 110 students. Among these students, 75 52 50 3 3 30 22 13 are are are are are are are taking taking taking taking taking taking taking English, history, math, English and history, English and math, history a nd math, English, history, and math. How many students are taking (a) English and history, but not math, (b) neither English, history, nor math, (c) math, but neither English nor history, (d) Engli sh, but not history, (e) only one of the three subjects, (f) exactly two of the three subjects. Problem 1.11 An experiment consists of the following two stages: (1) rst a fair die is rolled (2) if the number appearing is even, then a fair co in is tossed; if the number appearing is odd, then the die is tossed again. An o utcome of this experiment is a pair of the form (outcome from stage 1, outcome f rom stage 2). Let S be the collection of all outcomes. List the elements of S an d then nd the cardinality of S.
2 SET OPERATIONS 17 2 Set Operations In this section we introduce various operations on sets and study the properties of these operations. Complements If U is a given set whose subsets are under co nsideration, then we call U a universal set. Let U be a universal set and A, B b e two subsets of U. The absolute complement of A (See Figure 2.1(I)) is the set Ac = {x U x A}. Example 2.1 Find the complement of A = {1, 2, 3} if U = {1, 2, 3, 4, 5, 6}. Solution. From the de nition, Ac = {4, 5, 6} The relative complement of A with respect to B (See Figure 2.1(II)) is the set B A = {x U x B and x A}. Figure 2.1 Example 2.2 Let A = {1, 2, 3} and B = {{1, 2}, 3}. Find A B. Solution . The elements of A that are not in B are 1 and 2. That is, A B = {1, 2} Union a nd Intersection Given two sets A and B. The union of A and B is the set A B = {x x A or x B}
18 BASIC OPERATIONS ON SETS where the or is inclusive.(See Figure 2.2(a))
Figure 2.2 The above de nition can be extended to more than two sets. More precise ly, if A1 , A2 , , are sets then An = {x x Ai f or some i N}. n=1 The intersec n of A and B is the set (See Figure 2.2(b)) A B = {x x A and x B}. Example 2.3 E xpress each of the following events in terms of the events A, B, and C as well a s the operations of complementation, union and intersection: (a) at least one of the events A, B, C occurs; (b) at most one of the events A, B, C occurs; (c) no ne of the events A, B, C occurs; (d) all three events A, B, C occur; (e) exactly one of the events A, B, C occurs; (f) events A and B occur, but not C; (g) eith er event A occurs or, if not, then B also does not occur. In each case draw the corresponding Venn diagram. Solution. (a) A B C (b) (A B c C c ) (Ac B C c ) B c C) (Ac B c C c ) (c) (A B C)c = Ac B c C c (d) A B C (e) (A B C c ) (Ac B c C) (f) A B C c
2 SET OPERATIONS (g) A (Ac B c ) 19
Example 2.4 Translate the following set theoretic notation into event language. For example, A B means A or B occurs. (a) A B (b) A B (c) A B A B (d) A ( B (f) A B = Solution. (a) A and B occur (b) A occurs and B does not occur (c) A or B, but not both, occur (d) A occurs, and B and C do not occur (e) if A occurs , then B occurs (f) if A occurs, then B does not occur or if B occurs then A doe s not occur Example 2.5 Find a simpler expression of [(A B) (A C) (B c C c )] as suming all three sets intersect.
20 BASIC OPERATIONS ON SETS
Solution. Using a Venn diagram one can easily see that [(AB)(AC)(B c C c )] = A [A C)] If A B = we say that A and B are disjoint sets. Example 2.6 Let A and B be two non empty sets. Write A as the union of two disjoint sets. Solution. Using a Venn diagram one can easily see that A B and A B c are disjoint sets such that A = (A B) (A B c ) Example 2.7 Each team in a basketball league plays 20 games i n one tournament. Event A is the event that Team 1 wins 15 or more games in the tournament. Event B is the event that Team 1 wins less than 10 games. Event C is the event that Team 1 wins between 8 to 16 games. Of course, Team 1 can win at most 20 games. Using words, what do the following events represent? (a) A B and A B. (b) A C and A C. (c) B C and B C. (d) Ac , B c , and C c . Solution. (a) A B is the event that Team 1 wins 15 or more games or wins 9 or less games. A B is the empty set, since Team 1 cannot win 15 or more games and have less than 10 w ins at the same time. Therefore, event A and event B are disjoint. (b) A C is th e event that Team 1 wins at least 8 games. A C is the event that Team 1 wins 15 or 16 games. (c) B C is the event that Team 1 wins at most 16 games. B C is the event that Team 1 wins 8 or 9 games. (d) Ac is the event that Team 1 wins 14 or fewer games. B c is the event that Team 1 wins 10 or more games. C c is the even t that Team 1 wins fewer than 8 or more than 16 games
2 SET OPERATIONS 21
Given the sets A1 , A2 , , we de ne An = {x x Ai f or all i N}. n=1 Example 2. each positive integer n we de ne An = {n}. Find An . n=1 Solution. Clearly, An = n 1 Remark 2.1 Note that the Venn diagrams of A B and A B show that A B = B A and A B = B A. That is, and are commutative laws. The following theorem establishes the distributive laws of sets. Theorem 2.1 If A, B, and C are subsets of U then (a) A (B C) = (A B) (A C). (b) A (B C) = (A B) (A C). Proof. See Proble mark 2.2 Note that since and are commutative operations then (A B) C = (A C) (B C) and (A B) C = (A C) (B C). The following theorem presents the relationships b etween (A B)c , (A B)c , Ac and B c . Theorem 2.2 (De Morgans Laws) Let A and B b e subsets of U then (a) (A B)c = Ac B c . (b) (A B)c = Ac B c .
Proof. We prove part (a) leaving part(b) as an exercise for the reader. (a) Let x (A B)c . Then x U and x A B. Hence, x U and (x A and x B). This implies t x U and x A) and (x U and x B). It follows that x Ac B c . Conversely, let x B c . Then x Ac and x B c . Hence, x A and x B which implies that x (A B). Henc , x (A B)c Remark 2.3 De Morgans laws are valid for any countable number of sets. That is ( An )c = Ac n=1 n=1 n and ( An )c = Ac n=1 n=1 n Example 2.9 Let U be t t of people solicited for a contribution to a charity. All the people in U were given a chance to watch a video and to read a booklet. Let V be the set of peopl e who watched the video, B the set of people who read the booklet, C the set of people who made a contribution. (a) Describe with set notation: The set of people who did not see the video or read the booklet but who still made a contribution (b) Rewrite your answer using De Morgans law and and then restate the above. Solu tion. (a) (V B)c C. (b) (V B)c C = V c B c C = the set of people who did not wa ch the video, did not read the booklet, but did make a contribution If Ai Aj = f or all i = j then we say that the sets in the collection {An } are pairwise disjo int. n=1 Example 2.10 Find three sets A, B, and C that are not pairwise disjoint but A B C = . Solution. One example is A = B = {1} and C =
2 SET OPERATIONS Example 2.11 Find sets A1 , A2 , that are pairwise disjoint and An = . n=1 Solution. For each positive integer n, let An = {n} 23 Example 2.12 Throw a pair of fair dice. Let A be the event the total is 3, B the event the total is even, and C the event the total is a multiple of 7. Show tha t A, B, C are pairwise disjoint. Solution. We have A ={(1, 2), (2, 1)} B ={(1, 1 ), (1, 3), (1, 5), (2, 2), (2, 4), (2, 6)(3, 1), (3, 3), (3, 5), (4, 2), (4, 4), (4, 6), (5, 1), (5, 3), (5, 5), (6, 2), (6, 4), (6, 6)} C ={(1, 6), (2, 5), (3, 4), (4, 3), (5, 2), (6, 1)}. Clearly, A B = A C = B C = Next, we establish the following rule of counting. Theorem 2.3 (Inclusion Exclusion Principle) Suppose A and B are nite sets. Then (a) n(A B) = n(A) + n(B) n(A B). (b) If A B = , then (A B) = n(A) + n(B). (c) If A B, then n(A) n(B). Proof. (a) Indeed, n(A) gives t he number of elements in A including those that are common to A and B. The same holds for n(B). Hence, n(A) + n(B) includes twice the number of common elements. Therefore, to get an accurate count of the elements of A B, it is necessary to subtract n(A B) from n(A) + n(B). This establishes the result. (b) If A and B ar e disjoint then n(A B) = 0 and by (a) we have n(A B) = n(A) + n(B). (c) If A is a subset of B then the number of elements of A cannot exceed the number of eleme nts of B. That is, n(A) n(B)
Example 2.13 A total of 35 programmers interviewed for a job; 25 knew FORTRAN, 2 8 knew PASCAL, and 2 knew neither languages. How many knew both languages? Solut ion. Let F be the group of programmers that knew FORTRAN, P those who knew PASCA L. Then F P is the group of programmers who knew both languages. By the Inclusio n Exclusion Principle we have n(F P ) = n(F ) + n(P ) n(F P ). That is, 33 = 25 + 28 n(F P ). Solving for n(F P ) we nd n(F P ) = 20 Cartesian Product The notati on (a, b) is known as an ordered pair of elements and is de ned by (a, b) = {{a}, {a, b}}. The Cartesian product of two sets A and B is the set A B = {(a, b) a A, b B}. The idea can be extended to products of any number of sets. Given n sets A1 , A2 , , An the Cartesian product of these sets is the set A1 A2 An = a2 , , an ) : a1 A1 , a2 A2 , , an An } Example 2.14 Consider the experim tossing a fair coin n times. Represent the sample space as a Cartesian product. Solution. If S is the sample space then S = S1 S2 Sn , where Si , 1 i n, is set consisting of the two outcomes H=head and T = tail The following theorem is a tool for nding the cardinality of the Cartesian product of two nite sets. Theore m 2.4 Given two nite sets A and B. Then n(A B) = n(A) n(B).
2 SET OPERATIONS Proof. Suppose that A = {a1 , a2 , bm }. Then A B = {(a1 , b1 ), (a1 , b2 ), , , (a2 , bm ), (a3 , b1 ), (a3 , b2 ), , (a3 , , , (an , bm )} Thus, n(A B) = n m = n(A) outcomes of tossing a fair coin n times. 25
, an } and B = {b1 , b2 , , (a1 , bm ), (a2 , b1 ), (a2 , b2 ), bm ), . . . (an , b1 ), (an , b2 ) n(B) Example 2.15 What is the total o
Solution. If S is the sample space then S = S1 S2 Sn where Si , 1 i n, is t t consisting of the two outcomes H=head and T = tail. By the previous theorem, n (S) = 2n
Practice Problems Problem 2.1 Let A and B be any two sets. Use Venn diagrams to show that B = (A B ) (Ac B) and A B = A (Ac B). Problem 2.2 Show that if A B then B = A (Ac B) s, B can be written as the union of two disjoint sets. Problem 2.3 A survey of a groups viewing habits over the last year revealed the following information (i) (ii) (iii) (iv) (v) (vi) (vii) 28% watched gymnastics 29% watched baseball 19% w atched soccer 14% watched gymnastics and baseball 12% watched baseball and socce r 10% watched gymnastics and soccer 8% watched all three sports. Represent the statement the group that watched none of the three sports during th e last year using operations on sets. Problem 2.4 An urn contains 10 balls: 4 red and 6 blue. A second urn contains 16 red balls and an unknown number of blue ba lls. A single ball is drawn from each urn. For i = 1, 2, let Ri denote the event that a red ball is drawn from urn i and Bi the event that a blue ball is drawn from urn i. Show that the sets R1 R2 and B1 B2 are disjoint. Problem 2.5 An auto insurance has 10,000 policyholders. Each policyholder is classi ed as (i) young o r old; (ii) male or female; (iii) married or single.
2 SET OPERATIONS 27 Of these policyholders, 3000 are young, 4600 are male, and 7000 are married. The policyholders can also be classi ed as 1320 young males, 3010 married males, and 1400 young married persons. Finally, 600 of the policyholders are young married males. How many of the companys policyholders are young, female, and single? Prob lem 2.6 A marketing survey indicates that 60% of the population owns an automobi le, 30% owns a house, and 20% owns both an automobile and a house. What percenta ge of the population owns an automobile or a house, but not both? Problem 2.7 35 % of visits to a primary care physicians (PCP) o ce results in neither lab work no r referral to a specialist. Of those coming to a PCPs o ce, 30% are referred to sp ecialists and 40% require lab work. What percentage of visit to a PCPs o ce result s in both lab work and referral to a specialist? Problem 2.8 In a universe U of 100, let A and B be subsets of U such that n(A B) = 70 and n(A B c ) = 90. Deter mine n(A). Problem 2.9 An insurance company estimates that 40% of policyholders who have only an auto policy will renew next year and 60% of policyholders who h ave only a homeowners policy will renew next year. The company estimates that 80 % of policyholders who have both an auto and a homeowners policy will renew at l east one of those policies next year. Company records show that 65% of policyhol ders have an auto policy, 50% of policyholders have a homeowners policy, and 15% of policyholders have both an auto and a homeowners policy. Using the companys e stimates, calculate the percentage of policyholders that will renew at least one policy next year. Problem 2.10 Show that if A, B, and C are subsets of a univer se U then n(ABC) = n(A)+n(B)+n(C)n(AB)n(AC)n(BC)+n(ABC).
Problem 2.11 In a survey on the chewing gum preferences of baseball players, it was found that 22 like fruit. 25 like spearmint. 39 like grape. 9 like spearmint and fruit. 17 like fruit and grape. 20 like spearmint and grape. 6 like all avor s. 4 like none. How many players were surveyed? Problem 2.12 Let A, B, and C be three subsets of a universe U with the following properties: n(A) = 63, n(B) = 9 1, n(C) = 44, n(A B) = 25, n(A C) = 23, n(C B) = 21, n(A B C) = 139. Find n(A B C). Problem 2.13 In a class of students undergoing a computer course the followi ng were observed. Out of a total of 50 students: 30 know PASCAL, 18 know FORTRAN , 26 know COBOL, 9 know both PASCAL and FORTRAN, 16 know both PASCAL and COBOL, 8 know both FORTRAN and COBOL, 47 know at least one of the three languages. (a) How many students know none of these languages ? (b) How many students know all three languages ? Problem 2.14 Mr. Brown raises chickens. Each can be described as thin or fat, brown or red, hen or rooster. Four are thin brown hens, 17 are h ens, 14 are thin chickens, 4 are thin hens, 11 are thin brown chickens, 5 are br own hens, 3 are fat red roosters, 17 are thin or brown chickens. How many chicke ns does Mr. Brown have?
2 SET OPERATIONS 29
Problem 2.15 A doctor is studying the relationship between blood pressure and he artbeat abnormalities in her patients. She tests a random sample of her patients and notes their blood pressures (high, low, or normal) and their heartbeats (re gular or irregular). She nds that: (i) 14% have high blood pressure. (ii) 22% hav e low blood pressure. (iii) 15% have an irregular heartbeat. (iv) Of those with an irregular heartbeat, one third have high blood pressure. (v) Of those with no rmal blood pressure, one eighth have an irregular heartbeat. What portion of the patients selected have a regular heartbeat and low blood pressure? Problem 2.16 Prove: If A, B, and C are subsets of U then (a) A (B C) = (A B) (A C). (b) A ( C) = (A B) (A C). Problem 2.17 Translate the following verbal description of ev ents into set theoretic notation. For example, A or B occurs, but not both corresp onds to the set A B A B. (a) A occurs whenever B occurs. (b) If A occurs, then B does not occur. (c) Exactly one of the events A and B occurs. (d) Neither A nor B occur.
Counting and Combinatorics The major goal of this chapter is to establish several (combinatorial) technique s for counting large nite sets without actually listing their elements. These tec hniques provide e ective methods for counting the size of events, an important con cept in probability theory. 3 The Fundamental Principle of Counting Sometimes one encounters the question of listing all the outcomes of a certain e xperiment. One way for doing that is by constructing a so called tree diagram. E xample 3.1 A lottery allows you to select a two digit number. Each digit may be either 1,2 or 3. Use a tree diagram to show all the possible outcomes and tell h ow many di erent numbers can be selected. Solution. 31
32 COUNTING AND COMBINATORICS
The di erent numbers are {11, 12, 13, 21, 22, 23, 31, 32, 33} Of course, trees are manageable as long as the number of outcomes is not large. If there are many st ages to an experiment and several possibilities at each stage, the tree diagram associated with the experiment would become too large to be manageable. For such problems the counting of the outcomes is simpli ed by means of algebraic formulas . The commonly used formula is the Fundamental Principle of Counting which state s: Theorem 3.1 If a choice consists of k steps, of which the rst can be made in n 1 ways, for each of these the second can be made in n2 ways, , and for each of th ese the kth can be made in nk ways, then the whole choice can be made in n1 n2 n k ways. Proof. In set theoretic term, we let Si denote the set of outcomes for t he ith task, i = 1, 2, , k. Note that n(Si ) = ni . Then the set of outcomes for the entire job is the Cartesian product S1 S2 Sk = {(s1 , s2 , , sk ) : 1 i k}. Thus, we just need to show that n(S1 S2 Sk ) = n(S1 ) n(S2 ) e proof is by induction on k 2. Basis of Induction This is just Theorem 2.4. Ind uction Hypothesis Suppose n(S1 S2 Sk ) = n(S1 ) n(S2 ) n(Sk ). Inductio ion We must show n(S1 S2 Sk+1 ) = n(S1 ) n(S2 ) n(Sk+1 ). To see this, there is a one to one correspondence between the sets S1 S2 Sk+1 and (S1 S2 iven by f (s1 , s2 , , sk , sk+1 ) =
3 THE FUNDAMENTAL PRINCIPLE OF COUNTING 33
((s1 , s2 , , sk ), sk+1 ). Thus, n(S1 S2 Sk+1 ) = n((S1 S2 Sk 2 Sk )n(Sk+1 ) ( by Theorem 2.4). Now, applying the induction hypothesis gives n (S1 S2 Sk Sk+1 ) = n(S1 ) n(S2 ) n(Sk+1 ) Example 3.2 In designing a st e e ectiveness of migraine medicines, 3 factors were considered: (i) Medicine (A,B ,C,D, Placebo) (ii) Dosage Level (Low, Medium, High) (iii) Dosage Frequency (1,2 ,3,4 times/day) In how many possible ways can a migraine patient be given medici ne? Solution. The choice here consists of three stages, that is, k = 3. The rst s tage, can be made in n1 = 5 di erent ways, the second in n2 = 3 di erent ways, and t he third in n3 = 4 ways. Hence, the number of possible ways a migraine patient c an be given medecine is n1 n2 n3 = 5 3 4 = 60 di erent ways Example 3.3 How many l icense plates with 3 letters followed by 3 digits exist? Solution. A 6 step proc ess: (1) Choose the rst letter, (2) choose the second letter, (3) choose the thir d letter, (4) choose the rst digit, (5) choose the second digit, and (6) choose t he third digit. Every step can be done in a number of ways that does not depend on previous choices, and each license plate can be speci ed in this manner. So the re are 26 26 26 10 10 10 = 17, 576, 000 ways Example 3.4 How many numbers in the range 1000 9999 have no repeated digits? Solution. A 4 step process: (1) Choo se rst digit, (2) choose second digit, (3) choose third digit, (4) choose fourth digit. Every step can be done in a number of ways that does not depend on previo us choices, and each number can be speci ed in this manner. So there are 9 9 8 7 = 4, 536 ways
34 COUNTING AND COMBINATORICS Example 3.5 How many license plates with 3 letters followed by 3 digits exist if exactly one of the digits is 1? Solution. In this case, we must pick a place fo r the 1 digit, and then the remaining digit places must be populated from the di gits {0, 2, 9}. A 6 step process: (1) Choose the rst letter, (2) choose the secon d letter, (3) choose the third letter, (4) choose which of three positions the 1 goes, (5) choose the rst of the other digits, and (6) choose the second of the o ther digits. Every step can be done in a number of ways that does not depend on previous choices, and each license plate can be speci ed in this manner. So there are 26 26 26 3 9 9 = 4, 270, 968 ways
3 THE FUNDAMENTAL PRINCIPLE OF COUNTING 35 Practice Problems Problem 3.1 If each of the 10 digits is chosen at random, how many ways can you choose the following numbers? (a) A two digit code number, repeated digits permi tted. (b) A three digit identi cation card number, for which the rst digit cannot b e a 0. (c) A four digit bicycle lock number, where no digit can be used twice. ( d) A ve digit zip code number, with the rst digit not zero. Problem 3.2 (a) If eig ht horses are entered in a race and three nishing places are considered, how many nishing orders can they nish? Assume no ties. (b) If the top three horses are Luc ky one, Lucky Two, and Lucky Three, in how many possible orders can they nish? Pr oblem 3.3 You are taking 3 shirts(red, blue, yellow) and 2 pairs of pants (tan, gray) on a trip. How many di erent choices of out ts do you have? Problem 3.4 A club has 10 members. In how many ways can the club choose a president and vice presi dent if everyone is eligible? Problem 3.5 In a medical study, patients are class i ed according to whether they have blood type A, B, AB, or O, and also according to whether their blood pressure is low (L), normal (N), or high (H). Use a tree diagram to represent the various outcomes that can occur. Problem 3.6 If a trave l agency o ers special weekend trips to 12 di erent cities, by air, rail, or bus, in how many di erent ways can such a trip be arranged? Problem 3.7 If twenty paintin gs are entered in an art show, in how many di erent ways can the judges award a rst prize and a second prize?
36 COUNTING AND COMBINATORICS Problem 3.8 In how many ways can the 52 members of a labor union choose a presid ent, a vice president, a secretary, and a treasurer? Problem 3.9 Find the number of ways in which four of ten new movies can be ranked rst, second, third, and fo urth according to their attendance gures for the rst six months. Problem 3.10 How many ways are there to seat 10 people, consisting of 5 couples, in a row of seat s (10 seats wide) if all couples are to get adjacent seats?
4 PERMUTATIONS AND COMBINATIONS 37
4 Permutations and Combinations Consider the following problem: In how many ways can 8 horses nish in a race (ass uming there are no ties)? We can look at this problem as a decision consisting o f 8 steps. The rst step is the possibility of a horse to nish rst in the race, the second step is the possibility of a horse to nish second, , the 8th step is the p ossibility of a horse to nish 8th in the race. Thus, by the Fundamental Principle of Counting there are 8 7 6 5 4 3 2 1 = 40, 320 ways. This problem exhibits example of an ordered arrangement, that is, the order the objects are arranged i s important. Such an ordered arrangement is called a permutation. Products such as 8 7 6 5 4 3 2 1 can be written in a shorthand notation called factorial. T is, 8 7 6 5 4 3 2 1 = 8! (read 8 factorial). In general, we de ne n factor = n(n 1)(n 2) 3 2 1, n 1 where n is a whole number. By convention we de ne Example 4.1 Evaluate the following expressions: (a) 6! (b) 10! . 7! Solution. (a) 6! = 6 5 4 3 2 1 = 720 (b) 10! = 10987654321 orials we see that the number of permutations of n objects is n!. Example 4.2 ere are 6! permutations of the 6 letters of the word square. In how many of s r the second letter? Solution. Let r be the second letter. Then there are 5 ys to ll the rst spot, 4 ways to ll the third, 3 to ll the fourth, and so are 5! such permutations = 10 9 8 Th them i wa on. There
38 COUNTING AND COMBINATORICS Example 4.3 Five di erent books are on a shelf. In how many di erent ways could you arrange them? Solution. The ve books can be arranged in 5 4 3 2 1 = 5! = 120 ways Counting Permutations We next consider the permutations of a set of objects tak en from a larger set. Suppose we have n items. How many ordered arrangements of k items can we form from these n items? The number of permutations is denoted by P (n, k). The n refers to the number of di erent items and the k refers to the nu mber of them appearing in each arrangement. A formula for P (n, k) is given next . Theorem 4.1 For any non negative integer n and 0 k n we have P (n, k) = n! . ( n k)!
Proof. We can treat a permutation as a decision with k steps. The rst step can be made in n di erent ways, the second in n 1 di erent ways, ..., the k th in n k + 1 di erent ways. Thus, by the Fundamental Principle of Counting there are n(n 1) (n k + 1) kpermutations of n objects. That is, n! P (n, k) = n(n 1) (n k + 1) = n(n 1)(nk)! = (nk)! (nk)! Example 4.4 How many license plates are there that start with three letters followed by 4 digits (no repetitions)? Solution. The decision con sists of two steps. The rst is to select the letters and this can be done in P (2 6, 3) ways. The second step is to select the digits and this can be done in P (1 0, 4) ways. Thus, by the Fundamental Principle of Counting there are P (26, 3) P (10, 4) = 78, 624, 000 license plates Example 4.5 How many ve digit zip codes ca n be made where all digits are di erent? The possible digits are the numbers 0 thr ough 9.
4 PERMUTATIONS AND COMBINATIONS Solution. P (10, 5) = 39 10! (105)! = 30, 240 zip codes Circular permutations are ordered arrangements of objects in a circle. Suppose t hat circular permutations such as are considered as di erent. Consider seating n di erent objects in a circle under th e above assumption. Then each circular permutation corresponds to n linear permu tations (i.e. objects seated in a row) depending on where we start. Since there are exactly n! linear permutations, there are exactly n! = (n 1)! circular permu tations. However, if the two above permutations n are counted as the same, then the total number of circular permutations of this type is (n1)! . 2 Example 4.6 I n how many ways can you seat 6 persons at a circular dinner table. Solution. The re are (6 1)! = 5! = 120 ways to seat 6 persons at a circular dinner table. Co mbinations In a permutation the order of the set of objects or people is taken i nto account. However, there are many problems in which we want to know the numbe r of ways in which k objects can be selected from n distinct objects in arbitrar y order. For example, when selecting a two person committee from a club of 10 me mbers the order in the committee is irrelevant. That is choosing Mr. A and Ms. B in a committee is the same as choosing Ms. B and Mr. A. A combination is de ned a s a possible selection of a certain number of objects taken from a group without regard to order. More precisely, the number of
40 COUNTING AND COMBINATORICS kelement subsets of an nelement set is called the number of combinations of n obje cts taken k at a time. It is denoted by C(n, k) and is read n choose k. The formul a for C(n, k) is given next. Theorem 4.2 If C(n, k) denotes the number of ways i n which k objects can be selected from a set of n distinct objects then C(n, k) = n! P (n, k) = . k! k!(n k)! Proof. Since the number of groups of k elements out of n elements is C(n, k) and each group can be arranged in k! ways, we have P (n, k) = k!C(n, k). It follows that n! P (n, k) = C(n, k) = k! k!(n k)! An alternative notation for C(n, k) is or k > n. Example 4.7 From a group of 5 women and 7 men, how many di erent commit tees consisting of 2 women and 3 men can be formed? What if 2 of the men are feu ding and refuse to serve on the committee together? Solution. There are C(5, 2)C (7, 3) = 350 possible committees consisting of 2 women and 3 men. Now, if we sup pose that 2 men are feuding and refuse to serve together then the number of comm ittees that do not include the two men is C(7, 3) C(2, 2)C(5, 1) = 30 possible g roups. Because there are still C(5, 2) = 10 possible ways to choose the 2 women, it follows that there are 30 10 = 300 possible committees The next theorem disc usses some of the properties of combinations. Theorem 4.3 Suppose that n and k a re whole numbers with 0 k n. Then (a) C(n, 0) = C(n, n) = 1 and C(n, 1) = C(n, n 1) = n. (b) Symmetry property: C(n, k) = C(n, n k). (c) Pascals identity: C(n + 1, k) = C(n, k 1) + C(n, k). n k . We de ne C(n, k) = 0 if k < 0
4 PERMUTATIONS AND COMBINATIONS 41 Proof. n! (a) From the formula of C(, ) we have C(n, 0) = 0!(n0)! = 1 and C(n, n) = n! n! n! = 1. Similarly, C(n, 1) = 1!(n1)! = n and C(n, n 1) = (n1)! = n. n!(nn)! n! n! (b) Indeed, we have C(n, n k) = (nk)!(nn+k)! = k!(nk)! = C(n, k). (c) We have C(n, k 1) + C(n, k) = n! n! + (k 1)!(n k + 1)! k!(n k)! n!(n k + 1) n!k + = k!( n k + 1)! k!(n k + 1)! n! = (k + n k + 1) k!(n k + 1)! (n + 1)! = = C(n + 1, k) k!(n + 1 k)! Pascals identity allows one to construct the so called Pascals triangle (for n = 1 0) as shown in Figure 4.1. Figure 4.1 Example 4.8 The Chess Club has six members. In how many ways (a) can all six members line up for a picture? (b) can they choose a president and a sec retary? (c) can they choose three members to attend a regional tournament with n o regard to order?
42 COUNTING AND COMBINATORICS Solution. (a) P (6, 6) = 6! = 720 di erent ways (b) P (6, 2) = 30 ways (c) C(6, 3) = 20 di erent ways As an application of combination we have the following theorem which provides an expansion of (x + y)n , where n is a non negative integer. Theorem 4.4 (Binomial Theorem) Let x and y be variables, and let n be a non nega tive integer. Then n n (x + y) = k=0 C(n, k)xnk y k where C(n, k) will be called the binomial coe cient. Proof. The proof is by induction on n. Basis of induction: For n = 0 we have 0 (x + y)0 = k=0 C(0, k)x0k y k = 1. Induction hypothesis: Suppose that the theorem is true up to n. That is, n (x + y)n = k=0 C(n, k)xnk y k Induction step: Let us show that it is still true for n + 1. That is n+1 (x + y) n+1 = k=0 C(n + 1, k)xnk+1 y k .
4 PERMUTATIONS AND COMBINATIONS Indeed, we have (x + y)n+1 =(x + y)(x + y)n = x( x + y)n + y(x + y)n n n 43 =x k=0 n C(n, k)xnk y k + y k=0 n C(n, k)xnk y k C(n, k)xnk y k+1 = k=0 C(n, k)xnk+1 y k + k=0 =[C(n, 0)xn+1 + n, 1)xn1 y 2 + , 0)]xn y + + C(n + 1, 1)xn n+1 = k=0 C(n + 1, k)xnk+1 y k . Note that the coe cients in the expansion of (x + y)n are the entries of the (n + 1)st row of Pascals triangle. Example 4.9 Expand (x + y)6 using the Binomial Theo rem. Solution. By the Binomial Theorem and Pascals triangle we have (x + y)6 = x6 + 6x5 y + 15x4 y 2 + 20x3 y 3 + 15x2 y 4 + 6xy 5 + y 6 Example 4.10 How many su bsets are there of a set with n elements? Solution. Since there are C(n, k) subs ets of k elements with 0 k n, the total number of subsets of a set of n elements is n C(n, k) = (1 + 1)n = 2n k=0 C(n, 1)xn y + C(n, 2)xn1 y 2 + + C(n, n)xy n ] +[C(n, 0)xn y + C( + C(n, n 1)xy n + C(n, n)y n+1 ] =C(n + 1, 0)xn+1 + [C(n, 1) + C(n + [C(n, n) + C(n, n 1)]xy n + C(n + 1, n + 1)y n+1 =C(n + 1, 0)xn+1 y + C(n + 1, 2)xn1 y 2 + +C(n + 1, n)xy n + C(n + 1, n + 1)y n+1
44 COUNTING AND COMBINATORICS Practice Problems Problem 4.1 Find m and n so that P (m, n) = 9! 6! Problem 4.2 How many four letter code words can be formed using a standard 26 le tter alphabet (a) if repetition is allowed? (b) if repetition is not allowed? Pr oblem 4.3 Certain automobile license plates consist of a sequence of three lette rs followed by three digits. (a) If no repetitions of letters are permitted, how many possible license plates are there? (b) If no letters and no digits are rep eated, how many license plates are possible? Problem 4.4 A combination lock has 40 numbers on it. (a) How many di erent three number combinations can be made? (b) How many di erent combinations are there if the three numbers are di erent? Problem 4.5 (a) Miss Murphy wants to seat 12 of her students in a row for a class pictu re. How many di erent seating arrangements are there? (b) Seven of Miss Murphys stu dents are girls and 5 are boys. In how many di erent ways can she seat the 7 girls together on the left, and then the 5 boys together on the right? Problem 4.6 Us ing the digits 1, 3, 5, 7, and 9, with no repetitions of the digits, how many (a ) one digit numbers can be made? (b) two digit numbers can be made? (c) three di git numbers can be made? (d) four digit numbers can be made?
4 PERMUTATIONS AND COMBINATIONS 45 Problem 4.7 There are ve members of the Math Club. In how many ways can the posit ions of o cers, a president and a treasurer, be chosen? Problem 4.8 (a) A baseball team has nine players. Find the number of ways the manager can arrange the batt ing order. (b) Find the number of ways of choosing three initials from the alpha bet if none of the letters can be repeated. Name initials such as MBF and BMF ar e considered di erent. Problem 4.9 Find m and n so that C(m, n) = 13 Problem 4.10 The Library of Science Book Club o ers three books from a list of 42. If you circl e three choices from a list of 42 numbers on a postcard, how many possible choic es are there? Problem 4.11 At the beginning of the second quarter of a mathemati cs class for elementary school teachers, each of the classs 25 students shook han ds with each of the other students exactly once. How many handshakes took place? Problem 4.12 There are ve members of the math club. In how many ways can the two person Social Committee be chosen? Problem 4.13 A consumer group plans to select 2 televisions from a shipment of 8 to check the picture quality. In how many wa ys can they choose 2 televisions? Problem 4.14 A school has 30 teachers. In how many ways can the school choose 3 people to attend a national meeting? Problem 4 .15 Which is usually greater the number of combinations of a set of objects or t he number of permutations?
46 COUNTING AND COMBINATORICS Problem 4.16 How many di erent 12 person juries can be chosen from a rors? Problem 4.17 A jeweller has 15 di erent sized pearls to string band. In how many ways can this be done? Problem 4.18 Four teachers udents are seated in a circular discussion group. Find the number of an be done if teachers and students must be seated alternately. pool of 20 ju on a circular and four st ways this c
5 PERMUTATIONS AND COMBINATIONS WITH INDISTINGUISHABLE OBJECTS47
5 Permutations and Combinations with Indistinguishable Objects So far we have looked at permutations and combinations of objects where each obj ect was distinct (distinguishable). Now we will consider allowing multiple copie s of the same (identical/indistinguishable) item. Permutations with Repetitions( i.e. Indistinguishable Objects) We have seen how to calculate basic permutation problems where each element can be used one time at most. In the following discu ssion, we will see what happens when elements can be used repeatedly. As an exam ple, consider the following problem: How many di erent letter arrangements can be formed using the letters in DECEIVED? Note that the letters in DECEIVED are not all distinguishable since it contains repeated letters such as E and D. Thus, in terchanging the second and the fourth letters gives a result indistinguishable f rom interchanging the second and the seventh letters. The number of di erent rearr angement of the letters in DECEIVED can be found by applying the following theor em. Theorem 5.1 Given n objects of which n1 indistinguishable objects of type 1, n2 indistinguishable objects of type 2, , nk indistinguishable objects of type k where n1 + n2 + + nk = n. Then the number of distinguishable permutations of n objects is given by: n! . n1 !n2 ! nk ! Proof. The task of forming a permutatio n can be divided into k subtasks: For the rst kind we have n slots and n1 objects to place such that the order is unimportant. There are C(n, n1 ) di erent ways. F or the second kind we have n n1 slots and n2 objects to place such that the orde r is unimportant. There are C(nn1 , n2 ) ways. For the kth kind we have nn1 n2 nk s and nk objects to place such that the order is unimportant. There are C(n n1 n 2 nk1 , nk ) di erent ways. Thus, applying the Fundamental Principle of Counting nd that the number of distinguishable permutations is n! C(n, n1 ) C(n n1 , n2 ) C(n n1 nk1 , nk ) = n1 !n2 ! nk !
It follows that the number of di erent rearrangements of the letters in DE8! CEIVE D is 2!3! = 3360. Example 5.1 In how many ways can you arrange 2 red, 3 green, a nd 5 blue balls in a row? Solution. There are (2+3+5)! = 2520 di erent ways 2!3!5! Example 5.2 How many strings can be made using 4 As, 3 Bs, 7 Cs and 1 D? Solution. Applying the previous theorem with n = 4 + 3 + 7 + 1 = 15, n1 = 4, n2 = 3, n3 = 7, and n4 = 1 we nd that the total number of di erent such strings is 15! = 1, 801 , 800 7!4!3!1! Distributing Distinct Objects into Distinguishable Boxes We consi der the following problem: In how many ways can you distribute n distinct object s into k di erent boxes so that there will be n1 objects in Box 1, n2 in Box 2, , nk in Box k? where n1 + n2 + + nk = n. The answer to this question is provided b y the following theorem. Theorem 5.2 The number of ways to distribute n distinct objects into k distinguishable boxes so that there are ni objects placed into b ox i, 1 i k is: n! n1 !n2 ! nk ! Proof. The proof is a repetition of the proof o f Theorem 5.1: There are C(n, n1 ) possible choices for the rst box; for each cho ice of the rst box there are C(n n1 , n2 ) possible choices for the second box; f or each possible choice of the rst two boxes there are C(n n1 n2 ) possible choic es for the third box and so on. According to Theorem 3.1, there are C(n, n1 )C(nn1 , n2 )C(nn1 n2 , n3 ) C(nn1 n2 nk1 , nk ) = n! n1 !n2 ! di erent ways
5 PERMUTATIONS AND COMBINATIONS WITH INDISTINGUISHABLE OBJECTS49 Example 5.3 How many ways are there of assigning 10 police o cers to 3 di erent tasks: Patrol, of w hich there must be 5. O ce, of which there must be 2. Reserve, of which there will be 3. Solution. 10! There are 5!2!3! = 2520 di erent ways A correlation between Theorem 5.1 and Theorem 5.2 is shown next. Example 5.4 Solve Example 5.2 using Theorem 5 .2. Solution. Consider the string of length 15 composed of As, Bs, Cs, and Ds from E xample 5.2. Instead of placing the letters into the 15 unique positions, we plac e the 15 positions into the 4 boxes which represent the 4 letters. Now apply The orem 5.2 Example 5.5 Consider a group of 10 di erent people. Two committees are to be formed, one of four people and one of three people. (No person can serve on both committees.) In how many ways can the committees be formed. Solution. The t hree remaining people not chosen can be viewed as a third committee, so the number of choices is just the multinomial coe cient 10 4, 3, 3 = 10! = 4200 4!3!3! Distributing n Identical Objects into k Boxes Consider the following problem: Ho w many ways can you put 4 identical balls into 2 boxes? We will draw a picture i n order to understand the solution to this problem. Imagine that the 4 identical items are drawn as stars: { }.
50 COUNTING AND COMBINATORICS If we draw a vertical bar somewhere among these 4 stars, this can represent a un ique assignment of the balls to the boxes. For example: 0 Since there is a correspondence between a star/bar picture and assignments of ba lls to boxes, we can count the star/bar pictures and get an answer to our balls in boxes problem. So the question of nding the number of ways to put four identic al balls into 2 boxes is the same as the number of ways of arranging 5 5 four id entical stars and one vertical bar which is = = 5. This 4 1 is what one gets whe n replacing n = 4 and k = 2 in the following theorem. Theorem 5.3 The number of ways n identical (i.e. indistinguishable) objects can be distributed into k boxe s is n+k1 k1 = n+k1 n . Proof. Imagine the n identical objects as n stars. Draw k1 vertical e among these n stars. This can represent a unique assignment of the e boxes. Hence, there is a correspondence between a star/bar picture ents of balls to boxes. So how many ways can you arrange n identical 1 vertical bar? The answer is given by n+k1 k1 = n+k1 n bars somewher balls to th and assignm dots and k
It follows that there are C(n + k 1, k 1) ways of placing n identical objects in to k distinct boxes. Example 5.6 How many ways can we place 7 identical balls in to 8 separate (but distinguishable) boxes?
5 PERMUTATIONS AND COMBINATIONS WITH INDISTINGUISHABLE OBJECTS51 Solution. We ha ve n = 7, k = 8 so there are C(8 + 7 1, 7) di erent ways Remark 5.4 Note that Theo rem 5.3 gives the number of vectors (n1 , n2 , , nk ), where ni is a nonnegative integer, such that n1 + n2 + + nk = n where ni is the number of objects in box i, 1 i k. Example 5.7 How many solutions to x1 + x2 + x3 = 11 are there with xi nonnegative integers? Solution. This is like throwing 11 objects in 3 boxes. The objects inside each box are indistinguishable. Let xi be the number of objects in box i where i = 1, 2, 3. Then the number of solutions to the given equation i s 11 + 3 1 11 = 78. Example 5.8 How many solutions are there to the equation n1 + n2 + n3 + n4 = 21 where n1 2, n2 3, n3 4, and n4 5? Solution. We can rewrite the given equation in the form (n1 2) + (n2 3) + (n3 4) + (n4 5) = 7 or m1 + m2 + m3 + m4 = 7 where t he mi s are positive. The number of solutions is C(7+41, 7) = 120 Multinomial The orem We end this section by extending the binomial theorem to the multinomial th eorem when the original expression has more than two variables.
Theorem 5.4 For any variables x1 , x2 , , xr and any nonnegative integer n we ha ve (x1 + x2 + + xr )n = (n1 , n2 , , nr ) n1 + n2 + nr = n n! xn1 xn2 !n2 ! nr ! 1 2
Proof. We provide a combinatorial proof. A generic term in the expansion of (x1 + x2 + + xr )n will take the form Cxn1 xn2 xnr with n1 + n2 + nr = n. 2 1 eek a formula for the value of C. Obtaining a term of the form xn1 xn2 xnr invol ves chosing n1 terms of n 1 2 r terms of our expansion to be occupied by x1 , th en choosing n2 out of the remaining n n1 terms for x2 , choosing n3 of the remai ning n n1 n2 terms for x3 and so on. Thus, the number of times xn1 xn2 xnr occur in 1 2 r the expansion is C = C(n, n1 )C(n n1 , n2 )C(n n1 n2 , n3 ) C(n n1 , nr ) n! = n1 !n2 ! nr ! It follows from Remark 5.4 that the number of coe cients in the multinomial expansion is given by C(n + r 1, r 1). Example 5.9 What is t he coe cient of x3 y 6 z 12 in the expansion of (x + 2y 2 + 4z 3 )10 ? Solution. I f we expand using the trinomial theorem (three term version of the multinomial t heorem), then we have (x + 2y 2 + 4z 3 )10 = 10 n1 , n2 , n3 xn1 (2y 2 )n2 (4z 3 )n3 . where the sum is taken over all triples of numbers (n1 , n2 , n3 ) such that ni 0 and n1 + n2 + n3 = 10. If we want to nd the term containing x3 y 6 z 12 , we mu st choose n1 = 3, n2 = 3, and n3 = 4, and this term looks like 10 3, 3, 4 x3 (2y 2 )3 (4z 3 )4 = 10! 3 4 24 3!3!4! 10! 3 4 3 6 12 24xy z . 3!3!4! The coe cient of this term is
5 PERMUTATIONS AND COMBINATIONS WITH INDISTINGUISHABLE OBJECTS53 Example 5.10 Ho w many distinct 6 letter words can be formed from 3 As, 2 Bs, and a C? Express you r answer as a number. Solution. There are 6 3, 2, 1 = 6! = 60 3!2!1!
54 COUNTING AND COMBINATORICS Problems Problem 5.1 How many rearrangements of the word MATHEMATICAL are there? Problem 5.2 Suppose that a club with 20 members plans to form 3 distinct committees with 6, 5, and 4 members, respectively. In how many ways can this be done? Problem 5 .3 A deck of cards contains 13 clubs, 13 diamonds, 13 hearts and 13 spades, with 52 cards in all. In a game of bridge, 13 cards are dealt to each of 4 persons. In how many ways the 52 cards can be dealt evenly among 4 players? Problem 5.4 A book publisher has 3000 copies of a discrete mathematics book. How many ways ar e there to store these books in their three warehouses if the copies of the book are indistinguishable? Problem 5.5 A chess tournament has 10 competitors of whi ch 4 are from Russia, 3 from the United States, 2 from Great Britain, and 1 from Brazil. If the tournament result lists just the nationalities of the players in the order in which they placed, how many outcomes are possible? Problem 5.6 How many di erent signals, each consisting of 9 ags hung in a line, can be made from a set of 4 white ags, 3 red ags, and 2 blue ags if all ags of the same color are iden tical? Problem 5.7 An investor has $20,000 to invest among 4 possible investment s. Each investment must be a unit of $1,000. If the total $20,000 is to be inves ted, how many di erent investment strategies are possible? What if not all the mon ey need to be invested? Problem 5.8 How many solutions does the equation x1 + x2 + x3 = 5 have such that x1 , x2 , x3 are non negative integers?
5 PERMUTATIONS AND COMBINATIONS WITH INDISTINGUISHABLE OBJECTS55 Problem 5.9 (a) How many distinct terms occur in the expansion of (x + y + z)6 ? (b) What is th e coe cient of xy 2 z 3 ? (c) What is the sum of the coe cients of all the terms? Pr oblem 5.10 How many solutions are there to x1 + x2 + x3 + x4 + x5 = 50 for integ ers xi 1? Problem 5.11 A probability teacher buys 36 plain doughnuts for his cla ss of 25. In how many ways can the doughnuts be distributed? (Note: Not all stud ents need to have received a doughnut.) Problem 5.12 (a) How many solutions exis t to the equation x1 + x2 + x3 = 15, where x1 , x2 , and x3 have to be non negat ive integers? Simplify your answer as much as possible. [Note: the solution x1 = 12, x2 = 2, x3 = 1 is not the same as x1 = 1, x2 = 2, x3 = 12] (b) How many sol utions exist to the equation x1 x2 x3 = 36 213 , where x1 , x2 and x3 have to be positive integers? Hint: Let xi = 2ai 3bi , i = 1, 2, 3.
Probability: De nitions and Properties In this chapter we discuss the fundamental concepts of probability at a level at which no previous exposure to the topic is assumed. Probability has been used i n many applications ranging from medicine to business and so the study of probab ility is considered an essential component of any mathematics curriculum. So wha t is probability? Before answering this question we start with some basic de nitio ns. 6 Basic De nitions and Axioms of Probability An experiment is any situation whose outcomes cannot be predicted with certainty . Examples of an experiment include rolling a die, ipping a coin, and choosing a card from a deck of playing cards. By an outcome or simple event we mean any res ult of the experiment. For example, the experiment of rolling a die yields six o utcomes, namely, the outcomes 1,2,3,4,5, and 6. The sample space S of an experim ent is the set of all possible outcomes for the experiment. For example, if you roll a die one time then the experiment is the roll of the die. A sample space f or this experiment could be S = {1, 2, 3, 4, 5, 6} where each digit represents a face of the die. An event is a subset of the sample space. For example, the eve nt of rolling an odd number with a die consists of three simple events {1, 3, 5} . Example 6.1 Consider the random experiment of tossing a coin three times. 57
58 PROBABILITY: DEFINITIONS AND PROPERTIES (a) Find the sample space of this experiment. (b) Find the outcomes of the event of obtaining more than one head. Solution. We will use T for tail and H for hea d. (a) The sample space is composed of eight simple events: S = {T T T, T T H, T HT, T HH, HT T, HT H, HHT, HHH}. (b) The event of obtaining more than one head is the set {T HH, HT H, HHT, HHH} Probability is the measure of occurrence of an event. Various probability concepts exist nowadays. A widely used probability c oncept is the experimental probability which uses the relative frequency of an e vent and is de ned as follows. Let n(E) denote the number of times in the rst n rep etitions of the experiment that the event E occurs. Then P (E), the probability of the event E, is de ned by n(E) . n n This states that if we repeat an experiment a large number of times then the fraction of times the event E occurs will be cl ose to P (E). This result is a theorem called the law of large numbers which we will discuss in Section 39.1. The function P satis es the following axioms, known as Kolmogorov axioms: Axiom 1: For any event E, 0 P (E) 1. Axiom 2: P (S) = 1. A xiom 3: For any sequence of mutually exclusive events {En }n1 , that is Ei Ej = f or i = j, we have P (E) = lim P ( En ) = n=1 n=1 P (En ). (Countable Additivity) If we let E1 = S, En = ) = P (En ) = P (S) , E2 , , En } is k > n and Axioms 3 we n P (n Ek ) k=1 = k=1 P (Ek ).
for n > 1 then by Axioms 2 and 3 we have 1 = P (S) = P ( En + P (). This implies n=1 n=1 n=2 that P () = 0. Also, if {E1 a nite set of mutually exclusive events, then by de ning Ek = for nd
6 BASIC DEFINITIONS AND AXIOMS OF PROBABILITY 59 Any function P that satis es Axioms 1 3 will be called a probability measure. Ex ample 6.2 Consider the sample space S = {1, 2, 3}. Suppose that P ({1, 2}) = 0.5 and P ({2, 3}) = 0.7. Is P a valid probability measure? Justify your answer. So lution. We have P (1) + P (2) + P (3) = 1. But P ({1, 2}) = P (1) + P (2) = 0.5. This implies that 0.5 + P (3) = 1 or P (3) = 0.5. Similarly, 1 = P ({2, 3}) + P (1) = 0.7 + P (1) and so P (1) = 0.3. It follows that P (2) = 1 P (1) P (3) = 1 0.3 0.5 = 0.2. Since P (1) + P (2) + P (3) = 1, P is a valid probability measur e Example 6.3 If, for a given experiment, O1 , O2 , O3 , is an in nite sequence of outcomes, verify that i 1 P (Oi ) = , i = 1, 2, 3, 2 is a probability measure. Solution. Note that P (E) > 0 for any event E. Moreover, if S is the sample spac e then P (S) = i=1 P (Oi ) = 1 2 i=0 1 2 i = 1 1 2 1 1 2 =1 Now, if E1 , E2 , is a sequence of mutually exclusive events then P ( En ) n=1 = n=1 j=1 P (Onj ) = n=1 P (En ) where En = {On1 , On2 , }. Thus, P de nes a probability function Now, since E E c = S, E E c = , and P (S) = 1 we nd P (E c ) = 1 P (E). When the outcome of an expe riment is just as likely as another, as in the example of tossing a coin, the ou tcomes are said to be equally likely. The
60 PROBABILITY: DEFINITIONS AND PROPERTIES classical probability concept applies only when all possible outcomes are equall y likely, in which case we use the formula P (E) = number of outcomes favorable to event n(E) = , total number of outcomes n(S)
where n(E) denotes the number of elements in E. Since for any event E we have E S we can write 0 n(E) n(S) so that 0 n(E) 1. It follows that 0 P (E) 1. Clearly P (S) = 1. Also, n(S) Axiom 3 is easy to check using a generalization of Theore m 2.3 (b). Example 6.4 A hand of 5 cards is dealt from a deck. Let E be the even t that the hand contains 5 aces. List the elements of E. Solution. Recall that a standard deck of 52 playing cards can be described as follows: hearts (red) Ace 2 3 clubs (black) Ace 2 3 diamonds (red) Ace 2 3 spades (black) Ace 2 3 4 4 4 4 5 5 5 5 6 6 6 6 7 7 7 7 8 8 8 8 9 9 9 9 10 10 10 10 Jack Jack Jack Jack Queen Q ueen Queen Queen King King King King Cards labeled Ace, Jack, Queen, or King are called face cards. Since there are o nly 4 aces in the deck, event E is impossible, i.e. E = so that P (E) = 0 Exampl e 6.5 What is the probability of drawing an ace from a well shu ed deck of 52 play ing cards? Solution. Since there are four aces in a deck of 52 playing cards, th e probabiltiy of 1 4 getting an ace is 52 = 13 Example 6.6 What is the probabili ty of rolling a 3 or a 4 with a fair die? Solution. Since the event of having a 3 or a 4 has two simple events {3, 4}, the proba2 bility of rolling a 3 or a 4 i s 6 = 1 3
6 BASIC DEFINITIONS AND AXIOMS OF PROBABILITY 61
Example 6.7 In a room containing n people, calculate the chance that at least tw o of them have the same birthday. Solution. We have P(Two or more have birthday match) = 1 P(no birthday match) Since each person was born on one of the 365 d ays in the year, there are (365)n possible outcomes (assuming no one was born in Feb 29). Now, P(no birthday match) = Thus, P (Two or more have birthday match) =1 P (no birthday match) (365)(364) (365 n + 1) =1 (365)n Remark 6.5 It is impo tant to keep in mind that the above de nition of probability applies only to a sam ple space that has equally likely outcomes. Applying the de nition to a space with outcomes that are not equally likely leads to incorrect conclusions. For exampl e, the sample space for spinning the spinner in Figure 6.1 is given by S={Red,Bl ue}, but the outcome Blue is more likely to occur than is the outcome Red. Indee d, P (Blue) = 3 whereas P (Red) = 1 4 4 as supposed to P (Blue) = P (Red) = 1 2 (365)(364)(365n+1) (365)n Figure 6.1
62 PROBABILITY: DEFINITIONS AND PROPERTIES Practice Problems Problem 6.1 Consider the random experiment of rolling a die. (a) Find the sample space of this experiment. (b) Find the event of rolling the die an even number. Problem 6.2 An experiment consists of the following two stages: (1) rst a fair d ie is rolled (2) if the number appearing is even, then a fair coin is tossed; if the number appearing is odd, then the die is tossed again. An outcome of this e xperiment is a pair of the form (outcome from stage 1, outcome from stage 2). Le t S be the collection of all outcomes. Find the sample space of this experiment. Problem 6.3 An insurer o ers a health plan to the employees of a large company. A s part of this plan, the individual employees may choose exactly two of the supp lementary coverages A, B, and C, or they may choose no supplementary coverage. T he proportions of the companys employees that choose coverages 5 A, B, and C are 1 , 1 , and , 12 respectively. 4 3 Determine the probability that a randomly cho sen employee will choose no supplementary coverage. Problem 6.4 An experiment co nsists of throwing two four faced dice. (a) Write down the sample space of this experiment. (b) If E is the event total score is at least 4, list the outcomes bel onging to E c . (c) If each die is fair, nd the probability that the total score is at least 6 when the two dice are thrown. What is the probability that the tot al score is less than 6? (d) What is the probability that a double: (i.e. {(1, 1 ), (2, 2), (3, 3), (4, 4)}) will not be thrown? (e) What is the probability that a double is not thrown nor is the score greater than 6? Problem 6.5 Let S = {1, 2, 3, , 25}. If a number is chosen at random, that is, with the
6 BASIC DEFINITIONS AND AXIOMS OF PROBABILITY 63 same chance of being drawn as all other numbers in the set, calculate each of th e following probabilities: (a) The event A that an even number is drawn. (b) The event B that a number less than 10 and greater than 20 is drawn. (c) The event C that a number less than 26 is drawn. (d) The event D that a prime number is dr awn. (e) The event E that a number both even and prime is drawn. Problem 6.6 The following spinner is spun: Find the probabilities of obtaining each of the following: (a) P(factor of 35) ( b) P(multiple of 3) (c) P(even number) (d) P(11) (e) P(composite number) (f) P(n either prime nor composite) Problem 6.7 The game of bridge is played by four pla yers: north, south, east and west. Each of these players receive 13 cards. (a) W hat is the probability that one of the players receives all 13 spades? (b) Consi der a single hand of bridge. What is the probability for the hand to have exactl y 3 Aces? Problem 6.8 A cooler contains 15 cans of Coke and 10 cans of Pepsi. Th ree are drawn from the cooler without replacement. What is the probability that all three are the same brand? Problem 6.9 A coin is tossed repeatedly. What is t he probability that the second head
64 PROBABILITY: DEFINITIONS AND PROPERTIES appears at the 5th toss? (Hint: Since only the rst ve tosses matter, you can assum e that the coin is tossed only 5 times.) Problem 6.10 Suppose two nsided dice are rolled. De ne an appropriate probability space S and nd the probabilities of the f ollowing events: (a) the maximum of the two numbers rolled is less than or equal to 2; (b) the maximum of the two numbers rolled is exactly equal to 3. Problem 6.11 Suppose each of 100 professors in a large mathematics department picks at r andom one of 200 courses. What is the probability that two professors pick the s ame course? Problem 6.12 A fashionable club has 100 members, 30 of whom are lawy ers. 25 of the members are liars, while 55 are neither lawyers nor liars. (a) Ho w many of the lawyers are liars? (b) A liar is chosen at random. What is the pro bability that he is a lawyer? Problem 6.13 An inspector selects 2 items at rando m from a shipment of 5 items, of which 2 are defective. She tests the 2 items an d observes whether the sampled items are defective. (a) Write out the sample spa ce of all possible outcomes of this experiment. Be very speci c when identifying t hese. (b) The shipment will not be accepted if both sampled items are defective. What is the probability she will not accept the shipment?
7 PROPERTIES OF PROBABILITY 65
7 Properties of Probability In this section we discuss some of the important properties of P (). We de ne the p robability of nonoccurrence of an event E (called its failure) to be the number P (E c ). Since S = E E c and E E c = , P (S) = P (E) + P (E c ). Thus, P (E c ) = 1 P (E). Example 7.1 The probability that a college student without a u shot wi ll get the u is 0.45. What is the probability that a college student without the u shot will not get the u? Solution. Our sample space consists of those students w ho did not get the u shot. Let E be the set of those students without the u shot w ho did get the u. Then P (E) = 0.45. The probability that a student without the u shot will not get the u is then P (E c ) = 1 P (E) = 1 0.45 = 0.55 The union of t wo events A and B is the event A B whose outcomes are either in A or in B. The i ntersection of two events A and B is the event A B whose outcomes are outcomes o f both events A and B. Two events A and B are said to be mutually exclusive if t hey have no outcomes in common. In this case A B = and P (A B) = P () = 0. Exampl e 7.2 Consider the sample space of rolling a die. Let A be the event of rolling an even number, B the event of rolling an odd number, and C the event of rolling a 2. Find (a) A B, A C, and B C. (b) A B, A C, and B C. (c) Which events are m tually exclusive? Solution. (a) We have A B = {1, 2, 3, 4, 5, 6} A C = {2, 4, 6} B C = {1, 2, 3, 5}
66 (b) PROBABILITY: DEFINITIONS AND PROPERTIES AB = A C = {2} BC = (c) A and B are mutually exclusive as well as B and C Example 7. 3 Let A be the event of drawing a king from a well shu ed standard deck of playing c ards and B the event of drawing a ten card. Are A and B mutually exclusive? Soluti on. Since A = {king of diamonds, king of hearts, king of clubs, king of spades} and B = {ten of diamonds, ten of hearts, ten of clubs, ten of spades}, A and B a re mutually exclusive For any events A and B the probability of A B is given by the addition rule. Theorem 7.1 Let A and B be two events. Then P (A B) = P (A) + P (B) P (A B). Proof. Let Ac B denote the event whose outcomes are the outcomes in B that are not in A. Then using the Venn diagram below we see that B = (A B) (Ac B) and A B = A (Ac B).
7 PROPERTIES OF PROBABILITY 67 Since (A B) and (Ac B) are mutually exclusive, by Axiom 3 P (B) = P (A B) + P (A c B). Thus, P (Ac B) = P (B) P (A B). Similarly, A and Ac B are mutually exclusi ve, thus we have P (A B) = P (A) + P (Ac B) = P (A) + P (B) P (A B) Note that in the case A and B are mutually exclusive, P (A B) = 0 so that P (A B) = P (A) + P (B). Example 7.4 A mall has two elevetors. Let A be the event that the rst elev ator is busy, and let B be the event the second elevator is busy. Assume that P (A) = 0.2, P (B) = 0.3 and P (A B) = 0.06. Find the probability that neither of the elevators is busy. Solution. The probability that neither of the elevator is busy is P [(AB)c ] = 1P (A B). But P (A B) = P (A) + P (B) P (A B) = 0.2 + 0.3 6 = 0.44. Hence, P [(A B)c ] = 1 0.44 = 0.56 Example 7.5 Let P (A) = 0.9 and P ( B) = 0.6. Find the minimum possible value for P (A B). Solution. Since P (A) + P (B) = 1.5 and 0 P (A B) 1, by the previous theorem P (A B) = P (A) + P (B) P (A B) 1.5 1 = 0.5. So the minimum value of P (A B) is 0.5 Example 7.6 Suppose ther es 40% chance of colder weather, 10% chance of rain and colder weather, 80% chanc e of rain or colder weather. Find the chance of rain.
68 PROBABILITY: DEFINITIONS AND PROPERTIES
Solution. By the addition rule we have P (R) = P (R C) P (C) + P (R C) = 0.8 0.4 + 0.1 = 0.5 Example 7.7 Let N be the set of all positive integers and P be a pr obability measure de ned by P (n) = 32 for all n N. What is the probability that a number n chosen at random from N will be even? Solution. We have P ({2, 4, 6, } ) =P (2) + P (4) + P (6) + 2 2 2 = 2 + 4 + 6 + 3 3 3 2 1 1 = 2 1 + 2 + 4 + 3 1 1 = 1 + + 2 + 3 + 9 9 9 9 2 1 1 = 1 = 9 1 9 4 Finally, if E and F are two events such that E F, then F can be written as the union of two mutually exclusive eve nts F = E (E c F ). By Axiom 3 we obtain P (F ) = P (E) + P (E c F ). Thus, P (F ) P (E) = P (E c F ) 0 and this shows E F = P (E) P (F ). Theorem 7.2 For any t ree events A, B, and C we have P (A B C) =P (A) + P (B) + P (C) P (A B) P (A C) P (B C) +P (A B C).
7 PROPERTIES )) = P (A) + C) [P (A (A C) + P C) 69
OF PROBABILITY Proof. We have P (B) + P (C) P (B C) P B) + P (A C) P ((A B) ((A B) (A C)) = P (A) +
P (A B C) ((A B) (A (A C))] = P P (B) + P (C)
= P (A) + P (B C) P (A (B C)) = P (A) + P (B) + P (C) (A) + P (B) + P (C) P (B C) P (A B) P (A C) P (B
P C)
Example 7.8 If a person visits his dentist, suppose that the probability that he will have his teeth cleaned is 0.44, the probability that he will have a cavity lled is 0.24, the probability that he will have a tooth extracted is 0.21, the p robability that he will have his teeth cleaned and a cavity lled is 0.08, the pro bability that he will have his teeth cleaned and a tooth extracted is 0.11, the probability that he will have a cavity lled and a tooth extracted is 0.07, and th e probability that he will have his teeth cleaned, a cavity lled, and a tooth ext racted is 0.03. What is the probability that a person visiting his dentist will have at least one of these things done to him? Solution. Let C be the event that a person will have his teeth cleaned, F is the event that a person will have a cavity lled, and E is the event a person will have a tooth extracted. We are give n P (C) = 0.44, P (F ) = 0.24, P (E) = 0.21, P (C F ) = 0.08, P (C E) = 0.11, P (F E) = 0.07 and P (C F E) = 0.03. Thus, P (C F E) = 0.44 + 0.24 + 0.21 0.08 0.11 7 + 0.03 = 0.66
70 PROBABILITY: DEFINITIONS AND PROPERTIES Practice Problems Problem 7.1 If A and B are the events that a consumer testing service will rate a given stereo system very good or good, P (A) = 0.22, P (B) = 0.35. Find (a) P (Ac ); (b) P (A B); (c) P (A B). Problem 7.2 If the probabilities are 0.20, 0.15 , and 0.03 that a student will get a failing grade in Statistics, in English, or in both, what is the probability that the student will get a failing grade in a t least one of these subjects? Problem 7.3 If A is the event drawing an ace from a deck of cards and B is the event drawing a spade. Are A and B mutually exclusive? Find P (A B). Problem 7.4 A bag contains 18 colored marbles: 4 are colored red, 8 are colored yellow and 6 are colored green. A marble is selected at random. W hat is the probability that the marble chosen is either red or green? Problem 7. 5 Show that for any events A and B, P (A B) P (A) + P (B) 1. Problem 7.6 A golf bag contains 2 red tees, 4 blue tees, and 5 white tees. (a) What is the probabil ity of the event R that a tee drawn at random is red? (b) What is the probabilit y of the event not R that is, that a tee drawn at random is not red? (c) What is t he probability of the event that a tee drawn at random is either red or blue? Pr oblem 7.7 A fair pair of dice is rolled. Let E be the event of rolling a sum tha t is an even number and P the event of rolling a sum that is a prime number. Fin d the probability of rolling a sum that is even or prime?
7 PROPERTIES OF PROBABILITY 71 Problem 7.8 If events A and B are from the same sample space, and if P(A)=0.8 an d P(B)=0.9, can events A and B be mutually exclusive? Problem 7.9 A survey of a groups viewing habits over the last year revealed the following information (i) ( ii) (iii) (iv) (v) (vi) (vii) 28% watched gymnastics 29% watched baseball 19% wa tched soccer 14% watched gymnastics and baseball 12% watched baseball and soccer 10% watched gymnastics and soccer 8% watched all three sports. Find the probability of the group that watched none of the three sports during t he last year. Problem 7.10 The probability that a visit to a primary care physic ians (PCP) o ce results in neither lab work nor referral to a specialist is 35% . O f those coming to a PCPs o ce, 30% are referred to specialists and 40% require lab work. Determine the probability that a visit to a PCPs o ce results in both lab wor k and referral to a specialist. Problem 7.11 You are given P (A B) = 0.7 and P ( A B c ) = 0.9. Determine P (A). Problem 7.12 Among a large group of patients rec overing from shoulder injuries, it is found that 22% visit both a physical thera pist and a chiropractor, whereas 12% visit neither of these. The probability tha t a patient visits a chiropractor exceeds by 14% the probability that a patient visits a physical therapist. Determine the probability that a randomly chosen me mber of this group visits a physical therapist. Problem 7.13 In modeling the num ber of claims led by an individual under an automobile policy during a three year period, an actuary makes the simplifying
72 PROBABILITY: DEFINITIONS AND PROPERTIES assumption that for all integers n 0, pn+1 = 1 pn , where pn represents the 5 pr obability that the policyholder les n claims during the period. Under this assump tion, what is the probability that a policyholder les more than one claim during the period? Problem 7.14 Near a certain exit of I 17, the probabilities are 0.23 and 0.24 that a truck stopped at a roadblock will have faulty brakes or badly w orn tires. Also, the probability is 0.38 that a truck stopped at the roadblock w ill have faulty brakes and/or badly worn tires. What is the probability that a t ruck stopped at this roadblock will have faulty brakes as well as badly worn tir es? Problem 7.15 A marketing survey indicates that 60% of the population owns an automobile, 30% owns a house, and 20% owns both an automobile and a house. Calc ulate the probability that a person chosen at random owns an automobile or a hou se, but not both.
8 PROBABILITY AND COUNTING TECHNIQUES 73 8 Probability and Counting Techniques The Fundamental Principle of Counting can be used to compute probabilities as sh own in the following example. Example 8.1 A quiz has 5 multiple choice questions . Each question has 4 answer choices, of which 1 is correct and the other 3 are incorrect. Suppose that you guess on each question. (a) How many ways are there to answer the 5 questions? (b) What is the probability of getting all 5 question s right? (c) What is the probability of getting exactly 4 questions right and 1 wrong? (d) What is the probability of doing well (getting at least 4 right)? Sol ution. (a) We can look at this question as a decision consisting of ve steps. The re are 4 ways to do each step so that by the Fundamental Principle of Counting t here are (4)(4)(4)(4)(4) = 1024 ways (b) There is only one way to answer each qu estion correctly. Using the Fundamental Principle of Counting there is (1)(1)(1) (1)(1) = 1 way to answer all 5 questions correctly out of 1024 possibilities. He nce, P(all 5 right) = 1 1024 (c) The following table lists all possible responses that involve exactly 4 righ t answers, R stands for right and W stands for a wrong answer Five Responses Num ber of ways to ll out the test WRRRR (3)(1)(1)(1)(1) = 3 RWRRR (1)(3)(1)(1)(1) = 3 RRWRR (1)(1)(3)(1)(1) = 3 RRRWR (1)(1)(1)(3)(1) = 3 RRRRW (1)(1)(1)(1)(3) = 3 So there are 15 ways out of the 1024 possible ways that result in 4 right answer s and 1 wrong answer so that
74 PROBABILITY: DEFINITIONS AND PROPERTIES P(4 right,1 wrong) = 15 1024 1.5% (d) At least 4 means you can get either 4 right and 1 wrong or all 5 right. Thus, P (at least 4 right) =P (4right, 1wrong) + P (5 right) 15 1 = + 1024 1024 16 = 0 .016 1024 Example 8.2 1 Suppose S = {1, 2, 3, 4, 5, 6}. How many events A are th ere with P (A) = 3 ? Solution. We must have P ({i, j}) = 1 3 with i = j. There are C(6, 2) = 15 such events Probability Trees Probability trees can be used to compute the probabilities of combined outcomes in a sequence of experiments. Example 8.3 Construct the probab ility tree of the experiment of ipping a fair coin twice. Solution. The probabili ty tree is shown in Figure 8.1 Figure 8.1
8 PROBABILITY AND COUNTING TECHNIQUES 75 The probabilities shown in Figure 8.1 are obtained by following the paths leadin g to each of the four outcomes and multiplying the probabilities along the paths . This procedure is an instance of the following general property. Multiplicatio n Rule for Probabilities for Tree Diagrams For all multistage experiments, the p robability of the outcome along any path of a tree diagram is equal to the produ ct of all the probabilities along the path. Example 8.4 Suppose that out of 500 computer chips there are 9 defective. Construct the probability tree of the expe riment of sampling two of them without replacement. Solution. The probability tr ee is shown in Figure 8.2 Figure 8.2 Example 8.5 In a state assembly, 35% of the legislators are Democrats , and the other 65% are Republicans. 70% of the Democrats favor raising sales ta x, while only 40% of the Republicans favor the increase. If a legislator is sele cted at random from this group, what is the probability that he or she favors ra ising sales tax?
76 PROBABILITY: DEFINITIONS AND PROPERTIES Solution. Figure 8.3 shows a tree diagram for this problem. Figure 8.3 The rst and third branches correspond to favoring the tax. We add thei r probabilities. P (tax) = 0.245 + 0.26 = 0.505 Example 8.6 A regular insurance claimant is trying to hide 3 fraudulent claims among 7 genuine claims. The claim ant knows that the insurance company processes claims in batches of 5 or in batc hes of 10. For batches of 5, the insurance company will investigate one claim at random to check for fraud; for batches of 10, two of the claims are randomly se lected for investigation. The claimant has three possible strategies: (a) submit all 10 claims in a single batch, (b) submit two batches of 5, one containing 2 fraudulent claims and the other containing 1, (c) submit two batches of 5, one c ontaining 3 fraudulent claims and the other containing 0. What is the probabilit y that all three fraudulent claims will go undetected in each case? What is the best strategy? Solution. 7 7 (a) P(fraud not detected) = 10 6 = 15 9 (b) P(fraud not detected) = 3 4 = 12 5 5 25 (c) P(fraud not detected) = 2 1 = 2 5 5 Claiman ts best strategy is to split fraudulent claims between two batches of 5
8 PROBABILITY AND COUNTING TECHNIQUES 77 Practice Problems Problem 8.1 A box contains three red balls and two blue balls. Two balls are to be drawn without replacement. Use a tree diagram to represent the various outcom es that can occur. What is the probability of each outcome? Problem 8.2 Repeat t he previous exercise but this time replace the rst ball before drawing the second . Problem 8.3 A jar contains three red gumballs and two green gumballs. An exper iment consists of drawing gumballs one at a time from the jar, without replaceme nt, until a red one is obtained. Find the probability of the following events. A : Only one draw is needed. B : Exactly two draws are needed. C : Exactly three draws are needed. Problem 8.4 Consider a jar with three black marbles and one re d marble. For the experiment of drawing two marbles with replacement, what is th e probability of drawing a black marble and then a red marble in that order? Pro blem 8.5 A jar contains three marbles, two black and one red. Two marbles are dr awn with replacement. What is the probability that both marbles are black? Assum e that the marbles are equally likely to be drawn. Problem 8.6 A jar contains fo ur marbles one red, one green, one yellow, and one white. If two marbles are dra wn without replacement from the jar, what is the probability of getting a red ma rble and a white marble? Problem 8.7 A jar contains 3 white balls and 2 red ball s. A ball is drawn at random from the box and not replaced. Then a second ball i s drawn from the box. Draw a tree diagram for this experiment and nd the probabil ity that the two balls are of di erent colors.
78 PROBABILITY: DEFINITIONS AND PROPERTIES Problem 8.8 Suppose that a ball is drawn from the box in the previous problem, i ts color recorded, and then it is put back in the box. Draw a tree diagram for t his experiment and nd the probability that the two balls are of di erent colors. Pr oblem 8.9 Suppose there are 19 balls in an urn. They are identical except in col or. 16 of the balls are black and 3 are purple. You are instructed to draw out o ne ball, note its color, and set it aside. Then you are to draw out another ball and note its color. What is the probability of drawing out a black on the rst dr aw and a purple on the second? Problem 8.10 A class contains 8 boys and 7 girls. The teacher selects 3 of the children at random and without replacement. Calcul ate the probability that the number of boys selected exceeds the number of girls selected.
Conditional Probability and Independence In this chapter we introduce the concept of conditional probability. So far, the notation P (A) stands for the probability of A regardless of the occurrence of any other events. If the occurrence of an event B in uences the probability of A t hen this new probability is called conditional probability. 9 Conditional Probabilities We desire to know the probability of an event A conditional on the knowledge tha t another event B has occurred. The information the event B has occurred causes us to update the probabilities of other events in the sample space. To illustrat e, suppose you cast two dice; one red, and one green. Then the probability of ge tting two ones is 1/36. However, if, after casting the dice, you ascertain that the green die shows a one (but know nothing about the red die), then there is a 1/6 chance that both of them will be one. In other words, the probability of get ting two ones changes if you have partial information, and we refer to this (alt ered) probability as conditional probability. If the occurrence of the event A d epends on the occurrence of B then the conditional probability will be denoted b y P (A B), read as the probability of A given B. Conditioning restricts the samp le space to those outcomes which are in the set being conditioned on (in this ca se B). In this case, P (A B) = Thus, n(A B) P (A B) = = n(B) 79 n(AB) n(S) n(B) n(S) number of outcomes corresponding to event A and B . number of outcomes of B P (A B) P (B) =
80 CONDITIONAL PROBABILITY AND INDEPENDENCE provided that P (B) > 0. Example 9.1 Let A denote the event student is female and let B denote the event student is French. In a class of 100 students suppose 60 ar e French, and suppose that 10 of the French students are females. Find the proba bility that if I pick a French student, it will be a female, that is, nd P (A B). Solution. 10 Since 10 out of 100 students are both French and female, P (A B) = 100 = 60 0.1. Also, 60 out of the 100 students are French, so P (B) = 100 = 0.6 . Hence, 1 P (A B) = 0.1 = 6 0.6 From P (A B) = we can write P (A B) = P (A B)P (B) = P (B A)P (A). (9.1) P (A B) P (B) Example 9.2 Suppose an individual applying to a college determines that he has a n 80% chance of being accepted, and he knows that dormitory housing will only be provided for 60% of all of the accepted students. What is the probability that a student will be accepted and will receive dormitory housing? Solution. The pro bability of the student being accepted and receiving dormitory housing is de ned b y P(Accepted and Housing) = P(Housing Accepted)P(Accepted) = (0.6)(0.8) = 0.48
Equation (9.1) can be generalized to any nite number of events. Theorem 9.1 Consi der n events A1 , A2 , , An . Then P (A1 A2 An ) = P (A1 )P (A2 A1 )P (A3 A (An A1 A2 An1 )
9 CONDITIONAL PROBABILITIES Proof. We have P (A1 A2 An ) =P ((A1 A2 An1 ) An ) =P (An A1 A2 An 81
=P (An A1 A2 An1 )P (An1 A1 A2 An2 ) P (A1 A2 An A1 A2 An2 ) P (An2 A1 A2 An3 ) P (A2 A1 )P (A1 ) Writing the terms in reverse order we nd P (A1 A2 An ) = P (A1 )P (A2 A1 )P (A3 2 ) P (An A1 A2 An1 ) Example 9.3 Suppose 5 cards are drawn from a pack of Calculate the probability that all cards are the same suit, i.e. a ush. Solution . We must nd P(a ush) = P(5 spades) + P( 5 hearts ) + P( 5 diamonds ) + P( 5 clubs ) Now, the probability of getting 5 spades is found as follows: P(5 spades) = P (1st card is a spade)P(2nd card is a spade 1st card is a spade) P(5th card is a spade 1st,2nd,3rd,4th cards are spade) 9 13 = 52 12 11 10 48 51 50 49 Since the above calculation is the same for any of the four suits, P(a ush) = 4 13 52 12 51 11 50 10 49 9 48 We end this section by showing that P ( A) satis es the properties of ordinary prob abilities. Theorem 9.2 The function B P (B A) satis es the three axioms of ordinar y probabilities stated in Section 6.
82 CONDITIONAL PROBABILITY AND INDEPENDENCE
Proof. 1. Since 0 P (A B) P (A), 0 P (B A) 1. (SA) 2. P (S A) = PP (A) = P (A) = 1. P (A) 3. Suppose that B1 , B2 , , are mutually exclusive events. Then B1 A, B2 A, , are mutually exclusive. Thus, P ( Bn A) = n=1 = P ( (Bn A)) n=1 P (A) n=1 P (Bn A) = P (A) P (Bn A) n=1 Thus, every theorem we have proved for an ordinary probability function holds fo r a conditional probability function. For example, we have P (B c A) = 1 P (B A ). Prior and Posterior Probabilities The probability P (A) is the probability of the event A prior to introducing new events that might a ect A. It is known as th e prior probability of A. When the occurrence of an event B will a ect the event A then P (A B) is known as the posterior probability of A.
9 CONDITIONAL PROBABILITIES 83 Practice Problems Problem 9.1 A public health researcher examines the medical records of a group o f 937 men who died in 1999 and discovers that 210 of the men died from causes re lated to heart disease. Moreover, 312 of the 937 men had at least one parent who su ered from heart disease, and, of these 312 men, 102 died from causes related t o heart disease. Determine the probability that a man randomly selected from thi s group died of causes related to heart disease, given that neither of his paren ts su ered from heart disease. Problem 9.2 An insurance company examines its pool of auto insurance customers and gathers the following information: (i) All custo mers insure at least one car. (ii) 70% of the customers insure more than one car . (iii) 20% of the customers insure a sports car. (iv) Of those customers who in sure more than one car, 15% insure a sports car. Calculate the probability that a randomly selected customer insures exactly one car and that car is not a sport s car. Problem 9.3 An actuary is studying the prevalence of three health risk fa ctors, denoted by A, B, and C, within a population of women. For each of the thr ee factors, the probability is 0.1 that a woman in the population has only this risk factor (and no others). For any two of the three factors, the probability i s 0.12 that she has exactly these two risk factors (but not the other). The prob ability that a woman has all three risk factors, given that she has A and B, is 1 . 3 What is the probability that a woman has none of the three risk factors, g iven that she does not have risk factor A? Problem 9.4 3 You are given P (A) = 2 , P (A B) = 5 , P (B A) = 1 , P (C B) = 1 , and 5 4 3 P (C A B) = 1 . Find P (A B C). 2
84 CONDITIONAL PROBABILITY AND INDEPENDENCE Problem 9.5 The question, Do you smoke? was asked of 100 people. Results are shown in the table. Male Female Total Yes 19 12 31 No 41 28 69 Total 60 40 100 (a) What is the probability of a randomly selected individual being a male who s mokes? (b) What is the probability of a randomly selected individual being a mal e? (c) What is the probability of a randomly selected individual smoking? (d) Wh at is the probability of a randomly selected male smoking? (e) What is the proba bility that a randomly selected smoker is male? Problem 9.6 Suppose you have a b ox containing 22 jelly beans: 10 red, 5 green, and 7 orange. You pick two at ran dom. What is the probability that the rst is red and the second is orange? Proble m 9.7 Two fair dice are rolled. What is the (conditional) probability that the s um of the two faces is 6 given that the two dice are showing di erent faces? Probl em 9.8 A machine produces parts that are either good (90%), slightly defective ( 2%), or obviously defective (8%). Produced parts get passed through an automatic inspection machine, which is able to detect any part that is obviously defectiv e and discard it. What is the probability that a part that gets shipped will be of good quality? Problem 9.9 The probability that a person with certain symptoms has hepatitis is 0.8. A blood test used to con rm this diagnosis gives positive r esults for 90% of those who have the disease, and 5% of those who do not have th e disease. What is the probability that a person who reacts positively to the te st actually has hepatitis?
9 CONDITIONAL PROBABILITIES 85 Problem 9.10 If we randomly pick (without replacement) two television tubes in s uccession from a shipment of 240 television tubes of which 15 are defective, wha t is the probability that they will both be defective? Problem 9.11 Find the pro babilities of randomly drawing two aces in succession from an ordinary deck of 5 2 playing cards if we sample (a) without replacement (b) with replacement Proble m 9.12 A box of fuses contains 20 fuses, of which ve are defective. If three of t he fuses are selected at random and removed from the box in succession without r eplacement, what is the probability that all three fuses are defective? Problem 9.13 A study of drinking and driving has found that 40% of all fatal auto accide nts are attributed to drunk drivers, 1% of all auto accidents are fatal, and dru nk drivers are responsible for 20% of all accidents. Find the percentage of nonf atal accidents caused by drivers who are not drunk. Problem 9.14 A computer manu facturer buys cables from three rms. Firm A supplies 50% of all cables and has a 1% defective rate. Firm B supplies 30% of all cables and has a 2% defective rate . Firm C supplies the remaining 20% of cables and has a 5% defective rate. (a) W hat is the probability that a randomly selected cable that the computer manufact urer has purchased is defective (that is what is the overall defective rate of a ll cables they purchase)? (b) Given that a cable is defective, what is the proba bility it came from Firm A? From Firm B? From Firm C? Problem 9.15 In a certain town in the United States, 40% of the population are democrats and 60% are repub licans. The municipal government has proposed making gun ownership illegal in th e town. It is known that 75% of democrats and 30% of republicans support this me asure. If a resident of the town is selected at random
86 CONDITIONAL PROBABILITY AND INDEPENDENCE (a) what is the probability that they support the measure? (b) If the selected p erson does support the measure what is the probability the person is a democrat? (c) If the person selected does not support the measure, what is the probabilit y that he or she is a democrat?
10 POSTERIOR PROBABILITIES: BAYES FORMULA 87 10 Posterior Probabilities: Bayes Formula It is often the case that we know the probabilities of certain events conditiona l on other events, but what we would like to know is the reverse. That is, given P (A B) we would like to nd P (B A). Bayes formula is a simple mathematical formula used for calculating P (B A) given P (A B). We derive this formula as follows. Let A and B be two events. Then A = A (B B c ) = (A B) (A B c ). Since the event s A B and A B c are mutually exclusive, we can write P (A) = P (A B) + P (A B c ) = P (A B)P (B) + P (A B c )P (B c ) (10.1) Example 10.1 The completion of a construction job may be delayed because of a st rike. The probabilities are 0.60 that there will be a strike, 0.85 that the cons truction job will be completed on time if there is no strike, and 0.35 that the construction will be completed on time if there is a strike. What is the probabi lity that the construction job will be completed on time? Solution. Let A be the event that the construction job will be completed on time and B is the event th at there will be a strike. We are given P (B) = 0.60, P (A B c ) = 0.85, and P ( A B) = 0.35. From Equation (10.1) we nd P (A) = P (B)P (A B) + P (B c )P (A B c ) = (0.60)(0.35) + (0.4)(0.85) = 0.55 From Equation (10.1) we can get Bayes formul a: P (B A) = P (A B) P (A B)P (B) = . P (A) P (A B)P (B) + P (A B c )P (B c ) (1 0.2) Example 10.2 A company has two machines A and B for making shoes. It has been ob served that machine A produces 10% of the total production of shoes while machin e B produces 90% of the total production of shoes. Suppose that 1% of all the sh oes produced by A are defective while 5% of all the shoes produced by B are defe ctive. What is the probability that a shoe taken at random from a days production was made by the machine A, given that it is defective?
88 CONDITIONAL PROBABILITY AND INDEPENDENCE Solution. We are given P (A) = 0.1, P (B) = 0.9, P (D A) = 0.01, and P (D B) = 0 .05. We want to nd P (A D). Using Bayes formula we nd P (A D) = P (A D) P (D A)P (A ) = P (D) P (D A)P (A) + P (D B)P (B) (0.01)(0.1) 0.0217 = (0.01)(0.1) + (0.05)( 0.9) Example 10.3 A company that manufactures video cameras produces a basic model (B ) and a deluxe model (D). Over the past year, 40% of the cameras sold have been of the basic model. Of those buying the basic model, 30% purchased an extended w arranty, whereas 50% of all deluxe purchases do so. If you learn that a randomly selected purchaser has an extended warranty, how likely is it that he or she ha s a basic model? Solution. Let W denote the extended warranty. We are given P (B ) = 0.4, P (D) = 0.6, P (W B) = 0.3, and P (W D) = 0.5. By Bayes formula we hav e P (B W ) = P (W B)P (B) P (B W ) = P (W ) P (W B)P (B) + P (W D)P (D) (0.3) (0.4) = = 0.286 (0.3)(0.4) + (0.5)(0.6) Example 10.4 Approximately 1% of women aged 40 50 have breast cancer. A woman wi th breast cancer has a 90% chance of a positive test from a mammogram, while a w oman without breast cancer has a 10% chance of a false positive result. What is the probability a woman has breast cancer given that she just had a positive tes t? Solution. Let B = the woman has breast cancer and A = a positive result. We want to calculate P (B A) given by P (B A) = P (B A) . P (A)
10 POSTERIOR PROBABILITIES: BAYES FORMULA But P (B A) = P (A B)P (B) = (0.9)(0.01 ) = 0.009 and P (B c A) = P (A B c )P (B c ) = (0.1)(0.99) = 0.099. Thus, 9 0.00 9 = 0.009 + 0.099 108 Formula 10.2 is a special case of the more general result: P (B A) = 89 Theorem 10.1 (Bayes formula) Suppose that the sample space S is the union of mutu ally exclusive events H1 , H2 , , Hn with P (Hi ) > 0 for each i. Then for any e vent A and 1 i n we have P (A Hi )P (Hi ) P (Hi A) = P (A) where P (A) = P (H1 )P (A H1 ) + P (H2 )P (A H2 ) + + P (Hn )P (A Hn ). Proof. First note that P (A) =P (A S) = P (A (n Hi )) i=1 n =P (i=1 (A Hi )) n n = i=1 P (A Hi ) = i=1 P (A Hi )P (Hi ) Hence, P (Hi A) = P (A Hi )P (Hi ) = P (A) n i=1 P (A Hi )P (Hi ) P (A Hi )P (Hi ) Example 10.5 Suppose a voter poll is taken in three states. In state A, 50% of v oters support the liberal candidate, in state B, 60% of the voters support the l iberal candidate, and in state C, 35% of the voters support the liberal candidat e. Of the total population of the three states, 40% live in state A, 25% live in state B, and 35% live in state C. Given that a voter supports the liberal candi date, what is the probability that he/she lives in state B?
90 CONDITIONAL PROBABILITY AND INDEPENDENCE Solution. Let LI denote the event that a voter lives in state I, where I = A, B, or C. Let S denote the event that a voter supports liberal candidate. We want t o nd P (LB S). By Bayes formula we have P (LB S) = = P (S LB )P (LB ) P (S LA )P (LA )+P (S LB )P (LB )+P (S LC )P (LC ) (0.6)(0.25) 0.3175 (0.5)(0.4)+(0.6)(0.25)+(0.35)(0.35) Example 10.6 The members of a consulting rm rent cars from three rental companies : 60% from company 1, 30% from company 2, and 10% from company 3. Past statistic s show that 9% of the cars from company 1, 20% of the cars from company 2, and 6 % of the cars from company 3 need a tune up. If a rental car delivered to the rm needs a tune up, what is the probability that it came from company 2? Solution. De ne the events H1 H2 H3 T Then P (H1 ) = 0.6 P (H2 ) = 0.3 P (H3 ) = 0.1 P (T H 1 ) = 0.09 P (T H2 ) = 0.2 P (T H3 ) = 0.06 From Bayes theorem we have P (H2 T ) = P (T H2 )P (H2 ) P (T H1 )P (H1 ) + P (T H2 )P (H2 ) + P (T H3 )P (H3 ) 0.2 0.3 = = 0.5 0.09 0.6 + 0.2 0.3 + 0.06 0.1 =car =car =car =car comes from co mpany 1 comes from company 2 comes from company 3 needs tune up Example 10.7 The probability that a person with certain symptoms has hepatitis i s 0.8. A blood test used to con rm this diagnosis gives positive results for 90% o f those who have the disease, and 5% of those who do not have the disease. What is the probability that a person who reacts positively to the test actually has hepatitis ?
10 POSTERIOR PROBABILITIES: BAYES FORMULA 91 Solution. Let H denote the event that a person (with these symptoms) has Hepatit is, H c the event that they do not, and Y the event that the test results are po sitive. We have P (H) = 0.8, P (H c ) = 0.2, P (Y H) = 0.9, P (Y H c ) = 0.05. We want P (H Y ), and this is found by Bayes formula P (H Y ) = P (Y H)P (H) (0 .9)(0.8) 72 = = c )P (H c ) P (Y H)P (H) + P (Y H (0.9)(0.8) + (0.05)(0.2) 73 Example 10.8 A factory produces its products with three machines. Machine I, II, and III produces 50%, 30%, and 20% of the products, but 4%, 2%, and 4% of their products are defective, respectively. (a) What is the probability that a random ly selected product is defective? (b) If a randomly selected product was found t o be defective, what is the probability that this product was produced by machin e I? Solution. Let I, II, and III denote the events that the selected product is produced by machine I, II, and III, respectively. Let D be the event that the s elected product is defective. Then, P (I) = 0.5, P (II) = 0.3, P (III) = 0.2, P (D I) = 0.04, P (D II) = 0.02, P (D III) = 0.04. So, by the total probability ru le, we have P (D) =P (D I)P (I) + P (D II)P (II) + P (D III)P (III) =(0.04)(0.50 ) + (0.02)(0.30) + (0.04)(0.20) = 0.034 (b) By Bayes theorem, we nd P (I D) = (0.0 4)(0.50) P (D I)P (I) = 0.5882 P (D) 0.034 Example 10.9 A building has 60 occupants consisting of 15 women and 45 men. The men have probability 1/2 of being colorblind and the women have probability 1/3 of being colorblind. (a) Suppose you choose uniformly at random a person from th e 60 in the building. What is the probability that the person will be colorblind ? (b) Determine the conditional probability that you chose a woman given that th e occupant you chose is colorblind.
92 Solution. Let CONDITIONAL PROBABILITY AND INDEPENDENCE W ={the one selected is a woman} M ={the one selected is a man} B ={the one sele cted is colorblind} 1 (a) We are given the following information: P (W ) = 15 = 4 , P (M ) = 60 1 1 3 , P (B W ) = 3 , and P (B M ) = 2 . By the total law of probability we 4 have 11 P (B) = P (B W )P (W ) + P (B M )P (M ) = . 24 (b) Using Bayes theorem we nd P (W B) = P (B W )P (W ) (1/3)(1/4) 2 = = P (B) (11/24) 11
10 POSTERIOR PROBABILITIES: BAYES FORMULA 93 Practice Problems Problem 10.1 An insurance company believes that people can be divided into two c lasses: those who are accident prone and those who are not. Their statistics sho w that an accident prone person will have an accident at some time within 1 year period with probability 0.4, whereas this probability decreases to 0.2 for a no n accident prone person. (a) If we assume that 30% of the population is accident prone, what is the probability that a new policyholder will have an accident wi thin a year of purchasing a policy? (b) Suppose that a new policyholder has an a ccident within a year of purchasing a policy. What is the probability that he or she is accident prone? Problem 10.2 An auto insurance company insures drivers o f all ages. An actuary compiled the following statistics on the companys insured drivers: Age of Driver 16 20 21 30 31 65 66 99 Probability of Accident 0 .06 0.03 0.02 0.04 Portion of Companys Insured Drivers 0.08 0.15 0.49 0.28 A randomly selected driver that the company insures has an accident. Calculate t he probability that the driver was age 16 20. Problem 10.3 An insurance company issues life insurance policies in three separate categories: standard, preferred , and ultra preferred. Of the companys policyholders, 50% are standard, 40% are p referred, and 10% are ultra preferred. Each standard policyholder has probabilit y 0.010 of dying in the next year, each preferred policyholder has probability 0 .005 of dying in the next year, and each ultra preferred policyholder has probab ility 0.001 of dying in the next year. A policyholder dies in the next year. Wha t is the probability that the deceased policyholder was ultra preferred?
94 CONDITIONAL PROBABILITY AND INDEPENDENCE Problem 10.4 Upon arrival at a hospitals emergency room, patients are categorized according to their condition as critical, serious, or stable. In the past year: (i) (ii) (iii) (iv) (v) (vi) 10% of the emergency room patients were critical; 30% of the emergency room patients were serious; the rest of the emergency room patients were stable; 40% of the critical patients died; 10% of the serious pati ents died; and 1% of the stable patients died. Given that a patient survived, what is the probability that the patient was cate gorized as serious upon arrival? Problem 10.5 A health study tracked a group of persons for ve years. At the beginning of the study, 20% were classi ed as heavy sm okers, 30% as light smokers, and 50% as nonsmokers. Results of the study showed that light smokers were twice as likely as nonsmokers to die during the ve year s tudy, but only half as likely as heavy smokers. A randomly selected participant from the study died over the ve year period. Calculate the probability that the p articipant was a heavy smoker. Problem 10.6 An actuary studied the likelihood th at di erent types of drivers would be involved in at least one collision during an y one year period. The results of the study are presented below. Probability Typ e of Percentage of of at least one driver all drivers collision Teen 8% 0.15 You ng adult 16% 0.08 Midlife 45% 0.04 31% 0.05 Senior Total 100% Given that a drive r has been involved in at least one collision in the past year, what is the prob ability that the driver is a young adult driver?
10 POSTERIOR PROBABILITIES: BAYES FORMULA 95 Problem 10.7 A blood test indicates the presence of a particular disease 95% of the time when the disease is actually present. The same test indicates the prese nce of the disease 0.5% of the time when the disease is not present. One percent of the population actually has the disease. Calculate the probability that a pe rson has the disease given that the test indicates the presence of the disease. Problem 10.8 The probability that a randomly chosen male has a circulation probl em is 0.25 . Males who have a circulation problem are twice as likely to be smok ers as those who do not have a circulation problem. What is the conditional prob ability that a male has a circulation problem, given that he is a smoker? Proble m 10.9 A study of automobile accidents produced the following data: Model year 1 997 1998 1999 Other Proportion of all vehicles 0.16 0.18 0.20 0.46 Probability o f involvement in an accident 0.05 0.02 0.03 0.04 An automobile from one of the model years 1997, 1998, and 1999 was involved in a n accident. Determine the probability that the model year of this automobile is 1997. Problem 10.10 In a certain state, 25% of all cars emit excessive amounts o f pollutants. If the probability is 0.99 that a car emiting excessive amounts of pollutants will fail the states vehicular emission test, and the probability is 0.17 that a car not emitting excessive amounts of pollutants will nevertheless f ail the test, what is the probability that a car that fails the test actually em its excessive amounts of pollutants? Problem 10.11 I havent decided what to do ov er spring break. There is a 50% chance that
96 CONDITIONAL PROBABILITY AND INDEPENDENCE Ill go skiing, a 30% chance that Ill go hiking, and a 20% chance that Ill stay home and play soccer. The (conditional) probability of my getting injured is 30% if I go skiing, 10% if I go hiking, and 20% if I play soccer. (a) What is the proba bility that I will get injured over spring break? (b) If I come back from vacati on with an injury, what is the probability that I got it skiing? Problem 10.12 T he nal exam for a certain math class is graded pass/fail. A randomly chosen stude nt from a probability class has a 40% chance of knowing the material well. If he knows the material well, he has an 80% chance of passing the nal. If he doesnt kn ow the material well, he has a 40% chance of passing the nal anyway. (a) What is the probability of a randomly chosen student passing the nal exam? (b) If a stude nt passes, what is the probability that he knows the material? Problem 10.13 Ten percent of a companys life insurance policyholders are smokers. The rest are non smokers. For each nonsmoker, the probability of dying during the year is 0.01. F or each smoker, the probability of dying during the year is 0.05. Given that a p olicyholder has died, what is the probability that the policyholder was a smoker ? Problem 10.14 Suppose that 25% of all calculus students get an A, and that stu dents who had an A in calculus are 50% more likely to get an A in Math 408 as th ose who had a lower grade in calculus. If a student who received an A in Math 40 8 is chosen at random, what is the probability that he/she also received an A in calculus? (Assume all students in Math 408 have taken calculus.) Problem 10.15 A designer buys fabric from 2 suppliers. Supplier A provides 60% of the fabric, supplier B provides the remaining 40%. Supplier A has a defect rate of 5%. Suppl ier B has a defect rate of 8%. Given a batch of fabric is defective, what is the probability it came from Supplier A?
10 POSTERIOR PROBABILITIES: BAYES FORMULA 97 Problem 10.16 1/10 of men and 1/7 of women are color blind. A person is randomly selected. Assume males and females to be in equal numbers. (a) What is the prob ability that the selected person is color blind? (b) If the selected person is c olor blind, what is the probability that the person is male?
11 Independent Events Intuitively, when the occurrence of an event B has no in uence on the probability of occurrence of an event A then we say that the two events are indepedent. For example, in the experiment of tossing two coins, the rst toss has no e ect on the s econd toss. In terms of conditional probability, two events A and B are said to be independent if and only if P (A B) = P (A). We next introduce the two most ba sic theorems regarding independence. Theorem 11.1 A and B are independent events if and only if P (A B) = P (A)P (B). Proof. A and B are independent if and only if P (A B) = P (A) and this is equivalent to P (A B) = P (A B)P (B) = P (A)P (B ) Example 11.1 Show that P (A B) > P (A) if and only if P (Ac B) < P (Ac ). We assume that 0 < P (A) < 1 and 0 < P (B) < 1 Solution. We have P (A B) > P (A) P (A B) > P (A) P (B) P (A B) > P (A)P (B) P (B) P (A B) < P (B) P (A)P (B) P (A < P (B)(1 P (A)) P (Ac B) < P (B)P (Ac ) P (Ac B) < P (Ac ) P (B) P (Ac B) < P ( c )
11 INDEPENDENT EVENTS 99 Example 11.2 An oil exploration company currently has two active projects, one i n Asia and the other in Europe. Let A be the event that the Asian project is suc cessful and B the event that the European project is successful. Suppose that A and B are indepndent events with P (A) = 0.4 and P (B) = 0.7. What is the probab ility that at least one of the two projects will be successful? Solution. The pr obability that at least one of the two projects is successful is P (AB). Thus, P (A B) = P (A) + P (B) P (A B) = P (A) + P (B) P (A)P (B) = 0.82 Example 11.3 Let A and B be two independent events such that P (B A B) = P (A B) = 1 . What is P (B)? 2 Solution. First, note that P (A B) P (A)P (B) 1 = P (A B) = = = P (A). 2 P (B) P (B) Next, P (B AB) = Thus, 2 = 3 P (B) 1 2 2 3 and P (B) P (B) P (B) = = . P (A B) P (A) + P (B) P (A B) P (A) + P (B) P (A)P (B) + P (B) 2 1 2 Solving this equation for P (B) we nd P (B) =
Theorem 11.2 If A and B are independent then so are A and B c . Proof. First not e that A can be written as the union of two mutually exclusive
events: A = A (B B c ) = (A B) (A B c ). Thus, P (A) = P (A B) + P (A B c ). follows that P (A B c ) =P (A) P (A B) =P (A) P (A)P (B) =P (A)(1 P (B)) = P (A) P (B c ) Example 11.4 Show that if A and B are independent so are Ac and B c . S olution. Using De Morgans formula we have P (Ac B c ) =1 P (A B) = 1 [P (A) + P ( B) P (A B)] =[1 P (A)] P (B) + P (A)P (B) =P (Ac ) P (B)[1 P (A)] = P (Ac ) P )P (Ac ) =P (Ac )[1 P (B)] = P (Ac )P (B c ) Example 11.5 A probability for a wo man to have twins is 5%. A probability to win the lottery is 1:1,000,000. Joan h as twins. What is the probability for her to win the lottery? Solution. Since th e two events are independent, the probability for Joan to win the lottery is 1:1 ,000,000 When the outcome of one event a ects the outcome of a second event, the e vents are said to be dependent. The following is an example of events that are n ot independent. Example 11.6 Draw two cards from a deck. Let A = The rst card is a spade, and B = The second card is a spade. Show that A and B are dependent. Soluti on. We have P (A) = P (B) = 13 52 = 1 4 and 1 4 2 13 12 P (A B) = < 52 51 = P (A)P (B).
11 INDEPENDENT EVENTS By Theorem 11.1 the events A and B are dependent 101
The de nition of independence for a nite number of events is de ned as follows: Event s A1 , A2 , , An are said to be mutually independent or simply independent if fo r any 1 i1 < i2 < < ik n we have P (Ai1 Ai2 Aik ) = P (Ai1 )P (Ai2 ) particular, three events A, B, C are independent if and only if P (A B) =P (A)P (B) P (A C) =P (A)P (C) P (B C) =P (B)P (C) P (A B C) =P (A)P (B)P (C) Example 11.7 Consider the experiment of tossing a coin n times. Let Ai = the ith coin sho ws Heads. Show that A1 , A2 , , An are independent. Solution. For any 1 i1 < i2 < < ik n we have P (Ai1 Ai2 Aik ) = 21k . But P (Ai ) = 1 . Thus, P (Ai1 ) = P (Ai1 )P (Ai2 ) P (Aik ) 2 Example 11.8 The probability that a lab specimen contains high levels of contamination is 0.05. Four samples are checked, and th e samples are independent. (a) What is the probability that none contains high l evels of contamination? (b) What is the probability that exactly one contains hi gh levels of contamination? (c) What is the probability that at least one contai ns high levels of contamination? Solution. Let Hi denote the event that the ith sample contains high levels of contamination for i = 1, 2, 3, 4. The event that none contains high levels of contamination is equivalent to c c c c H1 H2 H3 H4 . So, by independence, the desired probability is c c c c c c c c P (H1 H2 H3 H4 ) =P (H1 )P (H2 )P (H3 )P (H4 ) =(1 0.05)4 = 0.81 45
102 (b) Let CONDITIONAL PROBABILITY AND INDEPENDENCE A1 A2 A3 A4
c c c =H1 H2 H3 H4 c c c =H1 H2 H3 H4 c c c =H1 H2 H3 H4 c c c =H1 H2 Then, the requested probability is the probability of the union A1 A2 A3 A4 and these events are mutually exclusive. Also, by independence, P (Ai ) = (0.95)3 (0 .05) = 0.0429, i = 1, 2, 3, 4. Therefore, the answer is 4(0.0429) = 0.1716. (c) Let B be the event that no sample contains high levels of contamination. The eve nt that at least one contains high levels of contamination is the complement of B, i.e. B c . By part (a), it is known that P (B) = 0.8145. So, the requested pr obability is P (B c ) = 1 P (B) = 1 0.8145 = 0.1855 Example 11.9 Find the probab ility of getting four sixes and then another number in ve random rolls of a balan ced die. Solution. Because the events are independent, the probability in questi on is 5 1 1 1 1 5 = 6 6 6 6 6 7776 A collection of events A1 , A2 , , An are d to be pairwise independent if and only if P (Ai Aj ) = P (Ai )P (Aj ) for any i = j where 1 i, j n. Pairwise independence does not imply independence as the f ollowing example shows. Example 11.10 The four sides of a tetrahedron (regular t hree sided pyramid with 4 sides consisting of isosceles triangles) are denoted b y 2, 3, 5 and 30, respectively. If the tetrahedron is rolled the number on the b ase is the outcome of interest. Consider the three events A =the number on the ba se is even, B =the number is divisible by 3, and C =the number is divisible by 5. Sho w that these events are pairwise independent, but not independent.
11 INDEPENDENT EVENTS Solution. Note that A = {2, 30}, B = {3, 30}, and C = {5, 40}. Then 1 1 1 = = P (A)P (B) 4 2 2 1 1 1 P (A C) =P ({30}) = = = P (A)P (C) 4 2 2 1 1 1 P (B C) =P ({30}) = = = P (B)P (C) 4 2 2 P (A B) =P ({30}) = 103 Hence, the events A, B, and C are pairwise independent. On the other hand P (A B C) = P ({30}) = 1 1 1 1 = = P (A)P (B)P (C) 4 2 2 2 so that A, B, and C are not independent
104 CONDITIONAL PROBABILITY AND INDEPENDENCE Practice Problems Problem 11.1 Determine whether the events are independent or dependent. (a) Sele cting a marble and then choosing a second marble without replacing the rst marble . (b) Rolling a number cube and spinning a spinner. Problem 11.2 David and Adria n have a coupon for a pizza with one topping. The choices of toppings are pepper oni, hamburger, sausage, onions, bell peppers, olives, and anchovies. If they ch oose at random, what is the probability that they both choose hamburger as a top ping? Problem 11.3 You randomly select two cards from a standard 52 card deck. W hat is the probability that the rst card is not a face card (a king, queen, jack, or an ace) and the second card is a face card if (a) you replace the rst card be fore selecting the second, and (b) you do not replace the rst card? Problem 11.4 You and two friends go to a restaurant and order a sandwich. The menu has 10 typ es of sandwiches and each of you is equally likely to order any type. What is th e probability that each of you orders a di erent type? Problem 11.5 An urn contain s 10 balls: 4 red and 6 blue. A second urn contains 16 red balls and an unknown number of blue balls. A single ball is drawn from each urn. The probability that both balls are the same color is 0.44 . Calculate the number of blue balls in t he second urn. Problem 11.6 An actuary studying the insurance preferences of aut omobile owners makes the following conclusions: (i) An automobile owner is twice as likely to purchase a collision coverage as opposed to a disability coverage. (ii) The event that an automobile owner purchases a collision coverage is indep endent of the event that he or she purchases a disability coverage.
11 INDEPENDENT EVENTS 105 (iii) The probability that an automobile owner purchases both collision and disa bility coverages is 0.15. What is the probability that an automobile owner purch ases neither collision nor disability coverage? Problem 11.7 An insurance compan y pays hospital claims. The number of claims that include emergency room or oper ating room charges is 85% of the total number of claims. The number of claims th at do not include emergency room charges is 25% of the total number of claims. T he occurrence of emergency room charges is independent of the occurrence of oper ating room charges on hospital claims. Calculate the probability that a claim su bmitted to the insurance company includes operating room charges. Problem 11.8 1 Let S = {1, 2, 3, 4} with each outcome having equal probability 4 and de ne the e vents A = {1, 2}, B = {1, 3}, and C = {1, 4}. Show that the three events are pai rwise independent but not independent. Problem 11.9 Assume A and B are independe nt events with P (A) = 0.2 and P (B) = 0.3. Let C be the event that neither A no r B occurs, let D be the event that exactly one of A or B occurs. Find P (C) and P (D). Problem 11.10 Suppose A, B, and C are mutually independent events with p robabilities P (A) = 0.5, P (B) = 0.8, and P (C) = 0.3. Find the probability tha t at least one of these events occurs. Problem 11.11 Suppose A, B, and C are mut ually independent events with probabilities P (A) = 0.5, P (B) = 0.8, and P (C) = 0.3. Find the probability that exactly two of the events A, B, C occur. Proble m 11.12 If events A, B, and C are independent, show that (a) A and B C are indep endent (b) A and B C are independent
106 CONDITIONAL PROBABILITY AND INDEPENDENCE Problem 11.13 Suppose you ip a nickel, a dime and a quarter. Each coin is fair, a nd the ips of the di erent coins are independent. Let A be the event the total value of the coins that came up heads is at least 15 cents. Let B be the event the quar ter came up heads. Let C be the event the total value of the coins that came up he ads is divisible by 10 cents. (a) Write down the sample space, and list the event s A, B, and C. (b) Find P (A), P (B) and P (C). (c) Compute P (B A). (d) Are B a nd C independent? Explain. Problem 11.14 Workplace accidents are categorized int o three groups: minor, moderate and severe. The probability that a given acciden t is minor is 0.5, that it is moderate is 0.4, and that it is severe is 0.1. Two accidents occur independently in one month. Calculate the probability that neit her accident is severe and at most one is moderate. Problem 11.15 Among undergra duate students living on a college campus, 20% have an automobile. Among undergr aduate students living o campus, 60% have an automobile. Among undergraduate stud ents, 30% live on campus. Give the probabilities of the following events when a student is selected at random: (a) Student lives o campus (b) Student lives on ca mpus and has an automobile (c) Student lives on campus and does not have an auto mobile (d) Student lives on campus and/or has an automobile (e) Student lives on campus given that he/she does not have an automobile.
12 ODDS AND CONDITIONAL PROBABILITY 107 12 Odds and Conditional Probability Whats the di erence between probabilities and odds? To answer this question, lets co nsider a game that involves rolling a die. If one gets the face 1 then he wins t he game, otherwise he loses. The probability of winning 5 is 1 whereas the proba bility of losing is 6 . The odds of winning is 1:5(read 6 1 to 5). This expressi on means that the probability of losing is ve times the probability of winning. T hus, probabilities describe the frequency of a favorable result in relation to a ll possible outcomes whereas the odds in favor compare the favorable outcomes to t he unfavorable outcomes. More formally, odds in favor = favorable outcomes unfavorable outcomes If E is the event of all favorable outcomes then its complementary, E c , is the event of unfavorable outcomes. Hence, odds in favor = n(E) n(E c ) Also, we de ne the odds against an event as odds against = unfavorable outcomes favorable outcomes = n(E c ) n(E) Any probability can be converted to odds, and any odds can be converted to a pro bability. Converting Odds to Probability Suppose that the odds for an event E is a:b. Thus, n(E) = ak and n(E c ) = bk where k is a positive integer. Since S = E E c and E E c = , by Theorem 2.3(b) we have n(S) = n(E) + n(E c ). Therefore, P (E) = and P (E c ) = n(E c ) n(S) n(E) n(S) = n(E) n(E)+n(E c ) = ak ak+bk = a a+b = n(E c ) n(E)+n(E c ) = bk ak+bk = b a+b Example 12.1 If the odds in favor of an event E is 5:4, compute P (E) and P (E c ). Solution. We have
108 CONDITIONAL PROBABILITY AND INDEPENDENCE P (E) = 5 5+4 = 5 9 and P (E c ) = 4 5+4 = 4 9 Converting Probability to Odds Given P (E), we want to nd the odds in favor of E and the odds against E. The odds in favor of E are n(E) n(E) n(S) = n(E c ) n(S) n(E c ) P (E) = P (E c ) P (E) = 1 P (E) and the odds against E are 1 P (E) n(E c ) = n(E) P (E) Example 12.2 For each of the following, nd the odds in favor of the events occurring: (a) Rolling a number less than 5 on a die. (b) Tossing hea ds on a fair coin. (c) Drawing an ace from an ordinary 52 card deck. Solution. ( a) The probability of rolling a number less than 5 is 4 and that of rolling 5 6 or 6 is 2 . Thus, the odds in favor of rolling a number less than 5 is 4 2 = 2 6 6 6 1 or 2:1 1 (b) Since P (H) = 1 and P (T ) = 2 , the odds in favor of gettin g heads is 2 1 1 2 or 1:1 2 4 (c) We have P(ace) = 52 and P(not an ace) = 48 so that the odds in favor of 52 1 4 drawing an ace is 52 48 = 12 or 1:12 52 Remark 12.1 A probability such as P (E) = 5 is just a ratio. The exact number of favora ble 6 outcomes and the exact total of all outcomes are not necessarily known. No w consider a hypothesis H that is true with probability P (H) and suppose that n ew evidence E is introduced. Then the conditional probabilities, given the evide nce E, that H is true and that H is not true are given by
12 ODDS AND CONDITIONAL PROBABILITY P (H E) = P (E H)P (H) P (E) 109 P (E H c )P (H c ) . P (E) and P (H c E) = Therefore, the new odds after the evidence E has been introduced is P (H E) P (H ) P (E H) = . c E) P (H P (H c ) P (E H c ) That is, the new value of the odds of H is its old value multiplied by the ratio of the conditional probability of the new evidence given that H is true to the conditional probability given that H is not true. Example 12.3 1 The probability that a coin C1 comes up heads is 4 and that of a coin C2 is 3 . Suppose that one of these coins is randomly chosen and is ipped twice. 4 If both ips land heads, what are the odds that coin C2 was the one ipped? Solution. Let H be the event that coin C2 was the one ipped and E t he event a coin ipped twice lands two heads. Since P (H) = P (H c ) P (H) P (E H) P (E H) P (H E) = = c E) c ) P (E H c ) P (H P (H P (E H c ) = 9 16 1 16 = 9. Hence, the odds are 9:1 that coin C2 was the one ipped
110 CONDITIONAL PROBABILITY AND INDEPENDENCE Practice Problems Problem 12.1 1 If the probability of a boy being born is 2 , and a family plans to have four children, what are the odds against having all boys? Problem 12.2 I f the odds against Deborahs winning rst prize in a chess tournament are 3:5, what is the probability that she will win rst prize? Problem 12.3 What are the odds in favor of getting at least two heads if a fair coin is tossed three times? Probl em 12.4 If the probability of rain for the day is 60%, what are the odds against its raining? Problem 12.5 On a tote board at a race track, the odds for Spiderm an are listed as 26:1. Tote boards list the odds that the horse will lose the ra ce. If this is the case, what is the probability of Spidermans winning the race? Problem 12.6 If a die is tossed, what are the odds in favor of the following eve nts? (a) Getting a 4 (b) Getting a prime (c) Getting a number greater than 0 (d) Getting a number greater than 6. Problem 12.7 Find the odds against E if P (E) = 3 . 4 Problem 12.8 Find P (E) in each case. (a) The odds in favor of E are 3:4 (b) The odds against E are 7:3 Problem 12.9 A nancial analyst states that based on his economic information, that the odds of a recession in the U.S. in the nex t two years is 2:1. State his beliefs in terms of the probability of a recession in the U.S. in the next two years.
Discrete Random Variables This chapter is one of two chapters dealing with random variables. After introdu cing the notion of a random variable, we discuss discrete random variables. Cont inuous random variables are left to the next chapter. 13 Random Variables By de nition, a random variable X is a function with domain the sample space and r ange a subset of the real numbers. For example, in rolling two dice X might repr esent the sum of the points on the two dice. Similarly, in taking samples of col lege students X might represent the number of hours per week a student studies, a students GPA, or a students height. The notation X(s) = x means that x is the va lue associated with the outcome s by the random variable X. There are three type s of random variables: discrete random variables, continuous random variables, a nd mixed random variables. A discrete random variable is usually the result of a count and therefore the range consists of whole integers. A continuous random v ariable is usually the result of a measurement. As a result the range is any sub set of the set of all real numbers. A mixed random variable is partially discret e and partially continuous. In this chapter we will just consider discrete rando m variables. Example 13.1 State whether the random variables are discrete or con tinuous. (a) A coin is tossed ten times. The random variable X is the number of tails that are noted. (b) A light bulb is burned until it burns out. The random variable Y is its lifetime in hours. 111
112 DISCRETE RANDOM VARIABLES Solution. (a) X can only take the values 0, 1, ..., 10, so X is a discrete rando m variable. (b) Y can take any positive real value, so Y is a continuous random variable Example 13.2 Toss a coin 3 times: S = {HHH, HHT, HT H, HT T, T HH, T HT , T T H, T T T }. Let X = # of Heads in 3 tosses. Find the range of X. Solution. We have X(HHH) = 3 X(HHT ) = 2 X(HT H) = 2 X(HT T ) = 1 X(T HH) = 2 X(T HT ) = 1 X(T T H) = 1 X(T T T ) = 0 Thus, the range of X consists of {0, 1, 2, 3} We us e upper case letters X, Y, Z, etc. to represent random variables. We use small l etters x, y, z, etc to represent possible values that the corresponding random v ariables X, Y, Z, etc. can take. The statement X = x de nes an event consisting of all outcomes with X measurement equal to x which is the set {s S : X(s) = x}. F or instance, considering the random variable of the previous example, the statem ent X = 2 is the event {HHT, HT H, T HH}. Because the value of a random variable i s determined by the outcomes of the experiment, we may assign probabilities to t he possible values of the random variable. For example, P (X = 2) = 3 . 8 Exampl e 13.3 Consider the experiment consisting of 2 rolls of a fair 4 sided die. Let X be a random variable, equal to the maximum of 2 rolls. Complete the following table x 1 2 P(X=x) Solution. x 1 1 P(X=x) 16 2 3 16 3 4 3 5 16 4 7 16
13 RANDOM VARIABLES 113 Example 13.4 Five male and ve female are ranked according to their scores on an e xamination. Assume that no two scores are alike and all 10! possible rankings ar e equally likely. Let X denote the highest ranking achieved by a female (for ins tance, X = 1 if the top ranked person is female). Find P (X = i), i = 1, 2, , 10 . Solution. Since 6 is the lowest possible rank attainable by the highest scorin g female, we must have P (X = 7) = P (X = 8) = P (X = 9) = P (X = 10) = 0. For X = 1 (female is highest ranking scorer), we have 5 possible choices out of 10 fo r the top spot that satisfy this requirement; hence P (X = 1) = 1 . 2 For X = 2 (female is 2nd highest scorer), we have 5 possible choices for the top male, the n 5 possible choices for the female who ranked 2nd overall, and then any arrange ment of the remaining 8 individuals is acceptable (out of 10! possible arrangeme nts of 10 individuals); hence, P (X = 2) = 5 5 5 8! = . 10! 18
For X = 3 (female is 3rd highest scorer), acceptable con gurations yield (5)(4)=20 possible choices for the top 2 males, 5 possible choices for the female who ran ked 3rd overall, and 7! di erent arrangement of the remaining 7 individuals (out o f a total of 10! possible arrangements of 10 individuals); hence, 5 5 4 5 7! = . P (X = 3) = 10! 36 Similarly, we have 5 5 4 3 5 6! = 10! 84 5 4 3 2 5 5! = 5) = = 10! 252 5 4 3 2 1 5 4! 1 P (X = 6) = = 10! 252 P (X = 4) =
114 DISCRETE RANDOM VARIABLES Practice Problems Problem 13.1 Determine whether the random variable is discrete or continuous. (a ) Time between oil changes on a car (b) Number of heart beats per minute (c) The number of calls at a switchboard in a day (d) Time it takes to nish a 60 minute exam Problem 13.2 Two socks are selected at random and removed in succession fro m a drawer containing ve brown socks and three green socks. List the elements of the sample space, the corresponding probabilities, and the corresponding values of the random variable X, where X is the number of brown socks selected. Problem 13.3 Suppose that two fair dice are rolled so that the sample space is S = {(i, j) : 1 i, j 6}. Let X be the random variable X(i, j) = i + j. Find P (X = 6). P roblem 13.4 Let X be a random variable with probability distribution table given below x 0 10 20 50 100 P(X=x) 0.4 0.3 0.15 0.1 0.05 Find P (X < 50). Problem 13 .5 Toss a fair coin until the coin lands on heads. Let the random variable X den ote the number of coin ips until the rst head appears. Find P (X = n) where n is a positive integer. Problem 13.6 Choose a baby name at random from S = { Josh, Jo hn, Thomas, Peter }. Let X() = rst letter in name. Find P (X = J). Problem 13.7 Th e number of injury claims per month is modeled by a random variable N ith 1 P ( N = n) = , n 0. (n + 1)(n + 2) Determine the probability of at least one claim d uring a particular month, given that there have been at most four claims during that month.
13 RANDOM VARIABLES 115 Problem 13.8 Let X be a discrete random variable with the following probability table x 1 5 10 50 100 P(X=x) 0.02 0.41 0.21 0.08 0.28 Compute P (X > 4 X 50). Pr oblem 13.9 A basketball player shoots three times, independently. The success pr obability of his rst, second, and third shots are 0.7, 0.5, and 0.4 respectively. Let X denote the number of successful shots among these three. (a) De ne the func tion X from the sample space S into R. (b) {X = 0} corresponds to what subset in S? What is the probability that he misses all three shots; i.e P (X = 0)? (c) { X = 1} corresponds to what subset in S? What is the probability that he succeeds exactly once among these three shots; i.e P (X = 1)? (d) {X = 2} corresponds to what subset in S? What is the probability that he succeeds exactly twice among these three shots; i.e P (X = 2)? (e) {X = 3} corresponds to what subset in S? W hat is the probability that he makes all three shots; i.e P (X = 3)? Problem 13. 10 Four distinct numbers are randomly distributed to players numbered 1 through 4. Whenever two players compare their numbers, the one with the higher one is de clared the winner. Initially, players 1 and 2 compare their numbers; the winner then compares with player 3, and so on. Let X denote the number of times player 1 is a winner. Find P (X = i), i = 0, 1, 2, 3. Problem 13.11 For a certain discr ete random variable on the non negative integers, the probability function satis e s the relationships P (0) = P (1), P (k + 1) = Find P (0). Problem 13.12 Under a n insurance policy, a maximum of ve claims may be led per year 1 P (k), k = 1, 2, 3, k
116 DISCRETE RANDOM VARIABLES by a policyholder. Let pn be the probability that a policyholder les n claims dur ing a given year, where n = 0, 1, 2, 3, 4, 5. An actuary makes the following obs ervations: (i) pn pn+1 for 0 n 4 (ii) The di erence between pn and pn+1 is the sam e for 0 n 4 (iii) Exactly 40% of policyholders le fewer than two claims during a given year. Calculate the probability that a random policyholder will le more tha n three claims during a given year.
14 PROBABILITY MASS FUNCTION AND CUMULATIVE DISTRIBUTION FUNCTION117 14 Probability Mass Function and Cumulative Distribution Function For a discrete random variable X, we de ne the probability distribution or the pro bability mass function by the equation p(x) = P (X = x). That is, a probability mass function (pmf) gives the probability that a discrete random variable is exa ctly equal to some value. The pmf can be an equation, a table, or a graph that s hows how probability is assigned to possible values of the random variable. Exam ple 14.1 Suppose a variable X can take the values 1, 2, 3, or 4. The probabiliti es associated with each outcome are described by the following table: x 1 2 3 4 p(x) 0.1 0.3 0.4 0.2 Draw the probability histogram. Solution. The probability h istogram is shown in Figure 14.1 Figure 14.1
118 DISCRETE RANDOM VARIABLES Example 14.2 A committee of 4 is to be selected from a group consisting of 5 men and 5 women. Let X be the random variable that represents the number of women i n the committee. Create the probability mass distribution. Solution. For x = 0, 1, 2, 3, 4 we have 5 x 5 4x 10 4 p(x) = . The probability mass function can be described by the table x 0 5 p(x) 210 1 50 210 2 100 210 3 50 210 4 5 210 Example 14.3 Consider the experiment of rolling a four sided die twice. Let X(1 , 2 ) = max{1 , 2 }. Find the equation of p(x). Solution. The pmf of X is p(x) = 2x1 16 0 2x 1 = I{1,2,3,4} (x) 16 if x = 1, 2, 3, 4 other ise
here I{1,2,3,4} (x) =
1 if x {1, 2, 3, 4} 0 other ise In general, e de ne the indicator function of a set A to be the function IA (x) = 1 if x A 0 other ise Note that if the range of a random variable is = {x1 , x2 , } then p(x) 0 p(x) = 0 ,x ,x
14 PROBABILITY MASS FUNCTION AND CUMULATIVE DISTRIBUTION FUNCTION119 Moreover, p (x) = 1. x All random variables (discrete and continuous) have a distribution function or c umulative distribution function, abbreviated cdf. It is a function giving the pr obability that the random variable X is less than or equal to x, for every value x. For a discrete random variable, the cumulative distribution function is foun d by summing up the probabilities. That is, F (a) = P (X a) = xa p(x). Example 14.4 Given the follo ing pmf p(x) = 1, if x = a 0, other ise Find a formula for F (x) and sketch its graph. Solution. A formula for F (x) is given by F (x) = Its graph is given in Figure 14.2 0, if x < a 1, other ise Figure 14.2 For discrete random variables the cumulative distribution function ill al ays be a step function ith jumps at each value of x that has probability greater than 0.
120 DISCRETE RANDOM VARIABLES Example 14.5 Consider the follo ing probability distribution x p(x) 1 2 3 4 0.25 0.5 0.125 0.125
Find a formula for F (x) and sketch its graph. Solution. The cdf is given by x<1 0 0.25 1 x < 2 0.75 2 x < 3 F (x) = 0.875 3 x < 4 1 4x Its grap ure 14.3 Figure 14.3 Note that the size of the step at any of the values 1,2,3,4 is equal to the probability that X assumes that particular value. That is, e have Theor em 14.1 If the range of a discrete random variable X consists of the values x1 < x2 < < xn then p(x1 ) = F (x1 ) and p(xi ) = F (xi ) F (xi1 ), i = 2, 3, ,
14 PROBABILITY MASS FUNCTION AND CUMULATIVE DISTRIBUTION FUNCTION121 Proof. Beca use F (x) = 0 for x < x1 then F (x1 ) = P (X x1 ) = P (X < x1 ) + P (X = x1 ) = P (X = x1 ) = p(x1 ). Also, by Proposition 21.7 of Section 21, for 2 i n we can write p(xi ) = P (xi1 < X xi ) = F (xi ) F (xi1 ) Example 14.6 If the distribution function of X is given by x<0 0 1 16 0 x < 1 5 1x<2 16 F (x) = 11 16 16 3 x < 4 1 x4 nd the pmf of X. Solution. Making use of the previous theorem, we get p(0) = 3 1 , p(3) = 1 , and p(4) = 16 and 0 otherwise 8 4 1 , p(1) 16 1 , p(2) 4 = =
122 DISCRETE RANDOM VARIABLES Practice Problems Problem 14.1 Consider the experiment of tossing a fair coin three times. Let X d enote the random variable representing the total number of heads. (a) Describe t he probability mass function by a table. (b) Describe the probability mass funct ion by a histogram. Problem 14.2 In the previous problem, describe the cumulativ e distribution function by a formula and by a graph. Problem 14.3 Toss a pair of fair dice. Let X denote the sum of the dots on the two faces. Find the probabil ity mass function. Problem 14.4 Take a carton of six machine parts, with one def ective and ve nondefective. Randomly draw parts for inspection, but do not replac e them after each draw, until a defective part fails inspection. Let X denote th e number of parts drawn until a nondefective part is found. Find the probability mass function of X and draw its histogram. Problem 14.5 Roll two dice, and let the random variable X denote the number of even numbers that appear. Find the pr obability mass function of X. Problem 14.6 Let X be a random variable with pmf 1 p(n) = 3 Find a formula for F (n). Problem 14.7 A box contains 100 computer chi ps, of which exactly 5 are good and the remaining 95 are bad. (a) Suppose rst you take out chips one at a time (without replacement) and 2 3 n , n = 0, 1, 2, .
14 PROBABILITY MASS FUNCTION AND CUMULATIVE DISTRIBUTION FUNCTION123 test each c hip you have taken out until you have found a good one. Let X be the number of c hips you have to take out in order to nd one that is good. (Thus, X can be as sma ll as 1 and as large as 96.) Find the probability distribution of X. (b) Suppose now that you take out exactly 10 chips and then test each of these 10 chips. Le t Y denote the number of good chips among the 10 you have taken out. Find the pr obability distribution of Y. Problem 14.8 Let X be a discrete random variable wi th cdf given by x < 4 0 3 4 x < 1 10 F (x) = 7 10 1 x < 4 1 x4 Find a x). Problem 14.9 A box contains 3 red and 4 blue marbles. Two marbles are random ly selected without replacement. If they are the same color then you win $2. If they are of di erent colors then you lose $ 1. Let X denote your gain/lost. Find t he probability mass function of X. Problem 14.10 An experiment consists of rando mly tossing a biased coin 3 times. The probability of heads on any particular to ss is known to be 1 . Let X denote 3 the number of heads. (a) Find the probabili ty mass function. (b) Plot the probability mass distribution function for X.
124 DISCRETE RANDOM VARIABLES
Problem 14.11 If the distribution function of X is given by x<2 0 1 36 2 x < 36 3 x < 4 6 36 4 x < 5 10 36 5 x < 6 15 6x<7 36 F (x) = < 9 30 36 9 x < 10 33 36 10 x < 11 35 36 11 x < 12 1 x tribution of X. Problem 14.12 A state lottery game draws 3 numbers at random (wi thout replacement) from a set of 15 numbers. Give the probability distribution f or the number of winning digits you will have when you purchase one ticket.
15 EXPECTED VALUE OF A DISCRETE RANDOM VARIABLE 125 15 Expected Value of a Discrete Random Variable A cube has three red faces, two green faces, and one blue face. A game consists of rolling the cube twice. You pay $2 to play. If both faces are the same color, you are paid $5(that is you win $3). If not, you lose the $2 it costs to play. Will you win money in the long run? Let W denote the event that you win. Then W = {RR, GG, BB} and 1 1 1 1 1 1 7 + + = 39%. 2 2 3 3 6 6 18 Thus, P (L) = 11 = 61 %. Hence, if you play the game 18 times you expect 18 to win 7 times and lose 11 times on average. So your winnings in dollars will be 3 7 2 11 = 1. That is, you can expect to lose $1 if you play the 1 game 18 times. On the average, you will lose $ 18 per game (about 6 cents). This can be found also using the equation P (W ) = P (RR) + P (GG) + P (BB) = 11 1 7 2 = 18 18 18 If we let X denote the winni ngs of this game then the range of X consists of the two numbers 3 and 2 which oc cur with respective probability 0.39 and 0.61. Thus, we can write 3 7 11 1 2 = . 18 18 18 We call this number the expected value of X. More formally, let the range of a discrete random variable X be a sequence of numbers x1 , x2 , , xk , and le t p(x) be the corresponding probability mass function. Then the expected value o f X is E(X) = 3 E(X) = x1 p(x1 ) + x2 p(x2 ) + + xk p(xk ). The following is a j usti cation of the above formula. Suppose that X has k possible values x1 , x2 , , xk and that pi = P (X = xi ) = p(xi ), i = 1, 2, , k. Suppose that in n repetit ions of the experiment, the number of times that X takes the value xi is ni . Th en the sum of the values of X over the n repetitions is n 1 x1 + n 2 x2 + + n k xk
126 and the average value of X is DISCRETE RANDOM VARIABLES
n1 x1 + n2 x2 + + nk xk n1 n2 nk = x1 + x2 + + xk . n n n n But P (X = xi ) = l mn ni . n Thus, the average value of X approaches E(X) = x1 p(x1 ) + x2 p(x2 ) + + xk p(xk ). The expected value of X is also know n as the mean value. Example 15.1 Suppose that an insurance company has broken d own yearly automobile claims for drivers from age 16 through 21 as shown in the following table. Amount of claim Probability $0 0.80 $ 2000 0.10 $ 4000 0.05 $ 6 000 0.03 $ 8000 0.01 $ 10000 0.01 How much should the company charge as its aver age premium in order to break even on costs for claims? Solution. Let X be the r andom variable of the amount of claim. Finding the expected value of X we have E (X) = 0(.80)+2000(.10)+4000(.05)+6000(.03)+8000(.01)+10000(.01) = 760 Since the average claim value is $760, the average automobile insurance premium should be set at $760 per year for the insurance company to break even Example 15.2 An Ame rican roulette wheel has 38 compartments around its rim. Two of these are colore d green and are numbered 0 and 00. The remaining compartments are numbered 1 thr ough 36 and are alternately colored black and red. When the wheel is spun in one direction, a small ivory ball is rolled in
15 EXPECTED VALUE OF A DISCRETE RANDOM VARIABLE 127 the opposite direction around the rim. When the wheel and the ball slow down, th e ball eventually falls in any one of the compartments with equal likelyhood if the wheel is fair. One way to play is to bet on whether the ball will fall in a red slot or a black slot. If you bet on red for example, you win the amount of t he bet if the ball lands in a red slot; otherwise you lose. What is the expected win if you consistently bet $5 on red? Solution. Let X denote the winnings of t he game. The probability of winning is and that of losing is 20 . Your expected win is 38 E(X) = 18 20 5 5 0.26 38 38 18 38 On average you should expect to lose 26 cents per play Example 15.3 Let A be a n onempty set. Consider the random variable I with range 0 and 1 and with pmf the indicator function IA where IA (x) = Find E(I). Solution. Since p(1) = P (A) and p(0) = p(Ac ), we have E(I) = 1 P (A) + 0 P (Ac ) = P (A) That is, the expected value of I is just the probability of A Example 15.4 A man determines that the probability of living 5 more years is 0.85. His insurance policy pays $1,000 if he dies within the next 5 years. Let X be the random variable that represents th e amount the insurance company pays out in the next 5 years. (a) What is the pro bability distribution of X? (b) What is the most he should be willing to pay for the policy? 1 if x A 0 otherwise
128 DISCRETE RANDOM VARIABLES Solution. (a) P (X = 1000) = 0.15 and P (X = 0) = 0.85. (b) E(X) = 1000 0.15 + 0 0.85 = 150. Thus, his expected payout is $150, so he should not be willing to p ay more than $150 for the policy Remark 15.2 The expected value (or mean) is rel ated to the physical property of center of mass. If we have a weightless rod in which weights of mass p(x) located at a distance x from the left endpoint of the rod then the point at which the rod is balanced is called the center of mass. I f is the center of m ss then we must h ve x (x )p(x) = 0. This equ tion implies t h t = x xp(x) = E(X). Thus, the expected v lue tells us something bout the cent er of the prob bility m ss function.
15 EXPECTED VALUE OF A DISCRETE RANDOM VARIABLE 129 Pr ctice Problems Problem 15.1 Compute the expected v lue of the sum of two f ces when rolling two dice. Problem 15.2 A g me consists of rolling two dice. You win the mounts sho wn for rolling the score shown. Score $ won 2 3 4 5 4 6 8 10 6 7 8 9 10 11 12 20 40 20 10 8 6 4 Compute the expected v lue of the g me. Problem 15.3 You pl y g me in which tw o dice re rolled. If sum of 7 ppe rs, you win $10; otherwise, you lose $2. I f you intend to pl y this g me for long time, should you expect to m ke money, lose money, or come out bout even? Expl in. Problem 15.4 Suppose it costs $8 t o roll p ir of dice. You get p id the sum of the numbers in doll rs th t ppe r on the dice. Wh t is the expected v lue of this g me? Problem 15.5 Consider th e spinner in Figure 15.1, with the p yo in e ch sector of the circle. Should the owner of this spinner expect to m ke money over n extended period of time if th e ch rge is $2.00 per spin? Figure 15.1
130 DISCRETE RANDOM VARIABLES Problem 15.6 An insur nce comp ny will insure your dorm room g inst theft for semester. Suppose the v lue of your possessions is $800. The prob bility of you r being 1 robbed of $400 worth of goods during semester is 100 , nd the prob bility 1 of your being robbed of $800 worth of goods is 400 . Assume th t these re the only possible kinds of robberies. How much should the insur nce comp ny ch rge people like you to cover the money they p y out nd to m ke n ddition l $20 pro t per person on the ver ge? Problem 15.7 Consider lottery g me in whic h 7 out of 10 people lose, 1 out of 10 wins $50, nd 2 out of 10 wins $35. If yo u pl yed 10 times, bout how much would you expect to win? Assume th t g me co sts doll r. Problem 15.8 You pick 3 di erent numbers between 1 nd 12. If you pi ck ll the numbers correctly you win $100. Wh t re your expected e rnings if it costs $1 to pl y? Problem 15.9 Two life insur nce policies, e ch with de th b ene t of 10,000 nd onetime premium of 500, re sold to couple, one for e ch p erson. The policies will expire t the end of the tenth ye r. The prob bility th t only the wife will survive t le st ten ye rs is 0.025, the prob bility th t only the husb nd will survive t le st ten ye rs is 0.01, nd the prob bility th t both of them will survive t le st ten ye rs is 0.96 . Wh t is the expected e xcess of premiums over cl ims, given th t the husb nd survives t le st ten ye r s? Problem 15.10 A professor h s m de 30 ex ms of which 8 re h rd, 12 re re son b le, nd 10 re e sy. The ex ms re mixed up nd the professor selects r ndomly four of them without repl cement to give to four sections of the course she is te ch ing. Let X be the number of sections th t receive h rd test. ( ) Wh t is the p rob bility th t no sections receive h rd test? (b) Wh t is the prob bility th t ex ctly one section receives h rd test? (c) Compute E(X).
15 EXPECTED VALUE OF A DISCRETE RANDOM VARIABLE Problem 15.11 Consider r ndom v ri ble X whose given by 0 0.2 0.5 F (x) = 0.6 0.6 + q 131
cumul tive distribution function is x < 2 2 x < 0 0 x < 2.2 2.2 x < 3 3x<4 x4 We are also told that P (X > 3) = 0.1. (a) What is q? (b) Compute P (X 2 > 2). ( c) What is p(0)? What is p(1)? What is p(P (X 0))? (Here, p(x) denotes the proba bility mass function (pmf) for X) (d) Find the formula of the function p(x). (e) Compute E(X). Problem 15.12 Used cars can be classi ed in two categories: peaches and lemons. Sellers know which type the car they are selling is. Buyers do not. Suppose that buyers are aware that the probability a car is a peach is 0.4. Sel lers value peaches at $2500, and lemons at $1500. Buyers value peaches at $3000, and lemons at $2000. (a) Obtain the expected value of a used car to a buyer who has no extra information. (b) Assuming that buyers will not pay more than their expected value for a used car, will sellers ever sell peaches? Problem 15.13 In a batch of 10 computer parts it is known that there are three defective parts. Four of the parts are selected at random to be tested. De ne the random variable X to be the number of working (non defective) computer parts selected. (a) Derive the probability mass function of X. (b) What is the cumulative distribution fun ction of X? (c) Find the expected value of X.
132 DISCRETE RANDOM VARIABLES 16 Expected Value of a Function of a Discrete Random Variable If we apply a function g() to a random variable X, the result is another ran1 dom variable Y = g(X). For example, X 2 , log X, X are all random variables derived from the original random variable X. In this section we are interested in nding the expected value of this new random variable. But rst we look at an example. Ex ample 16.1 Let X be a discrete random variable with range {1, 0, 1} and probabili ties P (X = 1) = 0.2, P (X = 0) = 0.5, and P (X = 1) = 0.3. Compute E(X 2 ). Solu tion. Let Y = X 2 . Then the range of Y is {0, 1}. Also, P (Y = 0) = P (X = 0) = 0.5 and P (Y = 1) = P (X = 1) + P (X = 1) = 0.2 + 0.3 = 0.5 Thus, E(X 2 ) = 0(0. 5)+1(0.5) = 0.5. Note that E(X) = 1(0.2)+0(0.5)+1(0.3) = 0.1 so that E(X 2 ) = (E (X))2 Now, if X is a discrete random variable and g(x) = x then we know that E(g (X)) = E(X) = xD xp(x) where D is the range of X and p(x) is its probability mass function. This sugges ts the following result for nding E(g(X)). Theorem 16.1 If X is a discrete random variable with range D and pmf p(x), then the expected value of any function g(X ) is computed by E(g(X)) = xD g(x)p(x). Proof. Let D be the range of X and D be the range of g(X). Thus, D = {g(x) : x D }.
16 EXPECTED VALUE OF A FUNCTION OF A DISCRETE RANDOM VARIABLE133 For each y D we let Ay = {x D : g(x) = y}. We will show that {s S : g(X)(s) = y} = xAy {s S : X(s ) = x}. The prove is by double inclusions. Let s S be such that g(X)(s) = y. The n g(X(s)) = y. Since X(s) D, there is an x D such that x = X(s) and g(x) = y. Th is shows that s xAy {s S : X(s) = x}. For the converse, let s xAy {s S : X(s) = Then there exists x D such that g(x) = y and X(s) = x. Hence, g(X)(s) = g(x) = y and this implies that s {s S : g(X)(s) = y}. Next, if x1 and x2 are two distinc t elements of Ay and w {s S : X(s) = x1 } {t S : X(t) = x2 } then this leads to x1 = x2 , a contradiction. Hence, {s S : X(s) = x1 } {t S : X(t) = x2 } = . From the above discussion we are in a position to nd pY (y), the pmf of Y = g(X), in t erms of the pmf of X. Indeed, pY (y) =P (Y = y) =P (g(X) = y) = xAy P (X = x) p(x) xAy = Now, from the de nition of the expected value we have E(g(X)) =E(Y ) = yD ypY (y) p(x) = yD y xAy = yD xAy g(x)p(x) g(x)p(x) xD = Note that the last equality follows from the fact that D is the disjoint unions of the Ay Example 16.2 If X is the number of points rolled with a balanced die, d the expected value of g(X) = 2X 2 + 1. n
134 DISCRETE RANDOM VARIABLES Solution. 1 Since each possible outcome has the probability 6 , we get 6 E[g(X)] = i=1 (2i2 + 1) 6 1 6 1 = 6 1 = 6 6+2 i=1 i2 = 94 3 6(6 + 1)(2 6 + 1) 6+2 6 As a consequence of the above theorem we have the following result. Corollary 16 .1 If X is a discrete random variable, then for any constants a and b we have E( aX + b) = aE(X) + b. Proof. Let D denote the range of X. Then E(aX + b) = xD (ax + b)p(x) xp(x) + b xD xD =a p(x) =aE(X) + b A similar argument establishes E(aX 2 + bX + c) = aE(X 2 ) + bE(X) + c. Example 16.3 Let X be a random variable with E(X) = 6 and E(X 2 ) = 45, and l et Y = 20 2X. Find E(Y ) and E(Y 2 ) [E(Y )]2 . Solution. By the properties of e xpectation, E(Y ) =E(20 2X) = 20 2E(X) = 20 12 = 8 E(Y 2 ) =E(400 80X + 4X 2 ) = 400 80E(X) + 4E(X 2 ) = 100
16 EXPECTED VALUE OF A FUNCTION OF A DISCRETE RANDOM VARIABLE135 E(Y 2 ) (E(Y )) 2 = 100 64 = 36 We conclude this section with the following de nition. If g(x) = x n then we call E(X n ) = x xn p(x) the nth moment of X. Thus, E(X) is the rst mom ent of X. Example 16.4 Show that E(X 2 ) = E(X(X 1)) + E(X). Solution. Let D be the range of X. We have E(X 2 ) = xD x2 p(x) (x(x 1) + x)p(x) xD = = xD x(x 1)p(x) + xD xp(x) = E(X(X 1)) + E(X) Remark 16.1 In our de nition of expectation the set D can be in nite. It is possible to have a random variable with unde ned expectation (See Problem 16.11).
136 DISCRETE RANDOM VARIABLES Practice Problems Problem 16.1 Suppose that X is a discrete random variable with probability mass function p(x) = cx2 , (a) Find the value of c. (b) Find E(X). (c) Find E(X(X 1)) . Problem 16.2 A random variable X has the following probability mass function d e ned in tabular form x 1 1 2 p(x) 2c 3c 4c (a) Find the value of c. (b) Compute p(1), p(1), and p(2). (c) Find E(X) and E(X 2 ). Problem 16.3 A die is loaded in such a way that the probability of any particular face showing is directly propo rtional to the number on that face. Let X denote the number showing for a single toss of this die. (a) What is the probability function p(x) for X? (b) What is the probability that an even number occurs on a single roll of this die? (c) Fin d the expected value of X. Problem 16.4 Let X be a discrete random variable. Sho w that E(aX 2 +bX +c) = aE(X 2 )+ bE(X) + c. Problem 16.5 Consider a random vari able X whose probability mass function is given by 0.1 x = 3 0.2 x=0 p(x) = x=3 0.1 0.3 x=4 0 otherwise x = 1, 2, 3, 4.
0.3 x
16 EXPECTED VALUE OF A FUNCTION OF A DISCRETE RANDOM VARIABLE137 Let F (x) be th e corresponding cdf. Find E(F (X)). Problem 16.6 An insurance policy pays 100 pe r day for up to 3 days of hospitalization and 50 per day for each day of hospita lization thereafter. The number of days of hospitalization, X, is a discrete ran dom variable with probability function p(k) = 6k 15 0 k = 1, 2, 3, 4, 5 otherwise
Determine the expected payment for hospitalization under this policy. Problem 16 .7 An insurance company sells a one year automobile policy with a deductible of 2 . The probability that the insured will incur a loss is 0.05 . If there is a l oss, the probability of a loss of amount N is K , for N = 1, , 5 and K N a const ant. These are the only possible loss amounts and no more than one loss can occu r. Determine the net premium for this policy. Problem 16.8 Consider a random var iable X whose probability mass function is given by 0.2 x = 1 0.3 x=0 0.1 p(x) = 0.1 x = 0.5 0.3 x=4 0 otherwise Find E(p(x)). Problem 16.9 A box cont s 3 red and 4 blue marbles. Two marbles are randomly selected without replacemen t. If they are the same color then you win $2. If they are of di erent colors then you lose $1. Let X denote the amount you win. (a) Find the probability mass fun ction of X. (b) Compute E(2X ).
138 DISCRETE RANDOM VARIABLES Problem 16.10 A movie theater sell three kinds of tickets children (for $3), a dult (for $8), and seniors (for $5). The number of children tickets has E[C] = 4 5. The number of adult tickets has E[A] = 137. Finally, the number of senior tic kets has E[S] = 34. You may assume the number of tickets in each group is indepe ndent. Any particular movie costs $300 to show, regardless of the audience size. (a) Write a formula relating C, A, and S to the theaters pro t P for a particular movie. (b) Find E(P ). Problem 16.11 If the probability distribution of X is giv en by p(x) = 1 2 x , x = 1, 2, 3, show that E(2X ) does not exist.
17 VARIANCE AND STANDARD DEVIATION 139 17 Variance and Standard Deviation In the previous section we learned how to nd the expected values of various funct ions of random variables. The most important of these are the variance and the s tandard deviation which give an idea about how spread out the probability mass f unction is about its expected value. The expected squared distance between the r andom variable and its mean is called the variance of the random variable. The p ositive square root of the variance is called the standard deviation of the rand om variable. If SD(X) denotes the standard deviation then the variance is given by the formula Var(X) = SD(X)2 = E [(X E(X))2 ] The variance of a random variabl e is typically calculated using the following formula Var(X) =E[(X E(X))2 ] =E[X 2 2XE(X) + (E(X))2 ] =E(X 2 ) 2E(X)E(X) + (E(X))2 =E(X 2 ) (E(X))2 where we hav e used the result of Problem 16.4. Example 17.1 We toss a fair coin and let X = 1 if we get heads, X = 1 if we get tails. Find the variance of X. Solution. 1 Sin ce E(X) = 1 2 1 Var(X) = 1 0 = 1 1 2 1 = 0 and E(X 2 ) = 12 2 + (1)2 1 2 = 1 we nd
A useful identity is given in the following result Theorem 17.1 If X is a discre te random variable then for any constants a and b we have Var(aX + b) = a2 Var(X )
140 DISCRETE RANDOM VARIABLES Proof. Since E(aX + b) = aE(X) + b we have Var(aX + b) =E (aX + b E(aX + b))2 =E [a2 (X E(X))2 ] =a2 E((X E(X))2 ) =a2 Var(X) Remark 17.1 Note that the units of Var(X) is the square of the units of X. This motivates the de nition of the standa rd deviation = Var(X) which i mea ured in the ame unit a X. Example 17.2 Thi year, Toronto Maple Leaf ticket co t an average of $80 with a variance of 10 5 quare dollar. Toronto city council want to charge a 3% tax on all ticket (i. e., all ticket will be 3% more expen ive). If thi happen , what would be the v ariance of the co t of Toranto Maple Leaf ticket ? Solution. Let X be the curre nt ticket price and Y be the new ticket price. Then Y = 1.03X. Hence, Var(Y ) = Var(1.03X) = 1.032 Var(X) = (1.03)2 (105) = 111.3945 Example 17.3 Roll one die a nd let X be the re ulting number. Find the variance and tandard deviation of X. Solution. We have E(X) = (1 + 2 + 3 + 4 + 5 + 6) and E(X 2 ) = (12 + 22 + 32 + 42 + 52 + 62 ) Thus, Var(X) = 35 91 49 = 6 4 12 1 21 7 = = 6 6 2 1 91 = . 6 6
17 VARIANCE AND STANDARD DEVIATION The standard deviation is SD(X) = 35 1.7078 1 2 141 Next, we will show that g(c) = E[(X c)2 ] is minimum when c = E(X). Theorem 17.2 Let c be a constant and let X be a random variable with mean E(X) and variance Var(X) < . Then (a) g(c) = E[(X c)2 ] = Var(X) + (c E(X))2 . (b) g(c) is minimize d at c = E(X). Proof. (a) We have E[(X c)2 ] =E[((X E(X)) (c E(X)))2 ] =E[(X E(X ))2 ] 2(c E(X))E(X E(X)) + (c E(X))2 =Var(X) + (c E(X))2 (b) Note that g (c) = 0 when c = E(X) and g (E(X)) = 2 > 0 so that g has a global minimum at c = E(X)
142 DISCRETE RANDOM VARIABLES Practice Problems Problem 17.1 A probability distribution of the claim sizes for an auto insurance policy is given in the table below: Claim size Probability 20 0.15 0.10 30 40 0 .05 50 0.20 0.10 60 70 0.10 80 0.30 What percentage of the claims are within one standard deviation of the mean claim size? Problem 17.2 A recent study indicate s that the annual cost of maintaining and repairing a car in a town in Ontario a verages 200 with a variance of 260. If a tax of 20% is introduced on all items a ssociated with the maintenance and repair of cars (i.e., everything is made 20% more expensive), what will be the variance of the annual cost of maintaining and repairing a car? Problem 17.3 A discrete random variable, X, has probability ma ss function p(x) = c(x 3)2 , x = 2, 1, 0, 1, 2. (a) Find the value of the constant c. (b) Find the mean and variance of X. Problem 17.4 In a batch of 10 computer parts it is known that there are three defective parts. Four of the parts are se lected at random to be tested. De ne the random variable X to be the number of wor king (non defective) computer parts selected. (a) Derive the probability mass fu nction of X. (b) Find the expected value and variance of X.
17 VARIANCE AND STANDARD DEVIATION 143 Problem 17.5 Suppose that X is a discrete random variable with probability mass function p(x) = cx2 , (a) Find the value of c. (b) Find E(X) and E(X(X 1)). (c) Find Var(X). Problem 17.6 Suppose X is a random variable with E(X) = 4 and Var(X ) = 9. Let Y = 4X + 5. Compute E(Y ) and Var(Y ). Problem 17.7 Many students com e to my o ce during the rst week of class with issues related to registration. The students either ask about STA200 or STA291 (never both), and they can either wan t to add the class or switch sections (again, not both). Suppose that 60% of the students ask about STA200. Of those who ask about STA200, 70% want to add the c lass. Of those who ask about STA291, 40% want to add the class. (a) Construct a probability table. (b) Of those who want to switch sections, what proportion are asking about STA291? (c) Suppose it takes 15 minutes to deal with STA200 and add ing, 12 minutes to deal with STA200 and switching, 8 minutes to deal with STA291 and adding, and 7 minutes to deal with STA291 and switching. Let X be the amount of ti me spend with a particular student. Find the probability mass function of X? d) What are the expectation and variance of X? Problem 17.8 A box contains 3 red an d 4 blue marbles. Two marbles are randomly selected without replacement. If they are the same color then you win $2. If they are of di erent colors then you loose $ 1. Let X denote the amount you win. (a) Find the probability mass function of X. (b) Compute E(X) and E(X 2 ). (c) Find Var(X). Problem 17.9 Let X be a discr ete random variable with probability mass function given in tabular form. x = 1, 2, 3, 4.
144 4 x 3 p(x) 10 DISCRETE RANDOM VARIABLES 1 4 10 4 3 10 Find the variance and the standard deviation of X. Problem 17.10 Let X be a rand om variable with probability distribution p(0) = 1p, p(1) = p, and 0 otherwise, w here 0 < p < 1. Find E(X) and V ar(X).
18 BINOMIAL AND MULTINOMIAL RANDOM VARIABLES 145 18 Binomial and Multinomial Random Variables Binomial experiments are problems that consist of a xed number of trials n, with each trial having exactly two possible outcomes: Success and failure. The probab ility of a success is denoted by p and that of a failure by q. Moreover, p and q are related by the formula p + q = 1. Also, we assume that the trials are indep endent, that is what happens to one trial does not a ect the probability of a succ ess in any other trial. The pre x bi in binomial experiment refers to the fact tha t there are two possible outcomes (e.g., head or tail, true or false, working or defective) to each trial in the binomial experiment. Example 18.1 Consider the experiment of tossing a fair coin. (a) What makes a trial? (b) What is a success ? a failure? Solution. (a) A trial consists of tossing the coin. (b) A success c ould be a Head and a failure is a Tail Let X represent the number of successes t hat occur in n trials. Then X is said to be a binomial random variable with para meters (n, p). If n = 1 then X is said to be a Bernoulli random variable. The ce ntral question of a binomial experiment is to nd the probability of r successes o ut of n trials. In the next paragraph well see how to compute such a probability. Now, anytime we make selections from a population without replacement, we do no t have independent trials. For example, selecting a ball from a box that contain balls of two di erent colors. Example 18.2 Privacy is a concern for many users of the Internet. One survey showed that 79% of Internet users are somewhat concern ed about the con dentiality of their e mail. Based on this information, we would l ike to nd the probability that for a random sample of 12 Internet users, 7 are co ncerned about the privacy of their e mail. What are the values of n, p, q, r?
146 DISCRETE RANDOM VARIABLES Solutions. This is a binomial experiment with 12 trials. If we assign success to an Internet user being concerned about the privacy of e mail, the probability o f success is 79%. We are interested in the probability of 7 successes. We have n = 12, p = 0.79, q = 1 0.79 = 0.21, r = 7 Binomial Distribution Formula As menti oned above, the central problem of a binomial experiment is to nd the probability of r successes out of n independent trials. In this section we see how to nd the se probabilities. Recall from Section 5 the formula for nding the number of combi nations of n distinct objects taken r at a time C(n, r) = n! . r!(n r)! Now, the probability of r successes in a sequence out of n independent trials is given by pr q nr . Since the binomial coe cient C(n, r) counts all the number of o utcomes that have r successes and nr failures, the probability of having r succes ses in any order is given by the binomial mass function p(r) = P (X = r) = C(n, r)pr q nr where p denotes the probability of a success and q = 1 p is the probabi lity of a failure. Note that by letting a = p and b = 1 p in the binomial formul a we nd n n p(k) = k=0 k=0 C(n, k)pk (1 p)nk = (p + 1 p)n = 1. Example 18.3 Suppose that in a particular sheet of 100 postage stamps, 3 are def ective. The inspection policy is to look at 5 randomly chosen stamps on a sheet and to release the sheet into circulation if none of those ve is defective. Write down the random variable, the corresponding probability distribution and then d etermine the probability that the sheet described here will be allowed to go int o circulation.
18 BINOMIAL AND MULTINOMIAL RANDOM VARIABLES 147 Solution. Let X be the number of defective stamps in the sheet. Then X is a bino mial random variable with probability distribution P (X = x) = C(5, x)(0.03)x (0 .97)5x , x = 0, 1, 2, 3, 4, 5. Now, P (sheet goes into circulation) = P (X = 0) = (0.97)5 = 0.859 Example 18.4 Suppose 40% of the student body at a large univers ity are in favor of a ban on drinking in dormitories. Suppose 5 students are to be randomly sampled. Find the probability that (a) 2 favor the ban. (b) less tha n 4 favor the ban. (c) at least 1 favor the ban. Solution. (a) P (X = 2) = C(5, 2)(.4)2 (0.6)3 0.3456. (b) P (X < 4) = P (0)+P (1)+P (2)+P (3) = C(5, 0)(0.4)0 ( 0.6)5 +C(5, 1)(0.4)1 (0.6)4 + C(5, 2)(0.4)2 (0.6)3 + C(5, 3)(0.4)3 (0.6)2 0.913. (c) P (X 1) = 1 P (X < 1) = 1 C(5, 0)(0.4)0 (0.6)5 0.922 Example 18.5 A tudent ha no knowledge of the material to be te ted on a true fal e exam with 10 que tion . So, the tudent ip a fair coin in order to determine the re pon e to each que tion. (a) What i the probability that the tudent an wer at lea t ix que tion correctly? (b) What i the probability that the tudent an wer at mo t t wo que tion correctly? Solution. (a) Let X be the number of correct re pon e . Then X i a binomial random variable with parameter n = 10 and p = 1 . So, the de ired probability i 2 P (X 6) =P (X = 6) + P (X = 7) + P (X = 8) + P (X = 9) + P (X = 10) 10 = x=6 C(10, x)(0.5)x (0.5)10x 0.3769.
148 (b) We have 2 DISCRETE RANDOM VARIABLES P (X 2) = x=0 C(10, x)(0.5)x (0.5)10x 0.0547
Example 18.6 A pediatrician reported that 30 percent of the children in the U.S. have above normal chole terol level . If thi i true, nd the probability that i n a ample of fourteen children te ted for chole terol, more than ix will have above normal chole terol level . Solution. Let X be the number of children in th e US with chole terol level above the normal. Then X i a binomial random variab le with n = 14 and p = 0.3. Thu , P (X > 6) = 1 P (X 6) = 1 P (X = 0) P (X = 1) P (X = 2) P (X = 3) P (X = 4) P (X = 5) P (X = 6) 0.0933 Next, we nd the expecte value and variation of a binomial random variable X. The expected value i foun d a follow . (n 1)! n! k pk (1 p)nk = np pk1 (1 p)nk E(X) = k!(n k)! (k 1)!(n k=1 k=1 n1 n n =np j=0 (n 1)! pj (1 p)n1j = np(p + 1 p)n1 = np j!(n 1 j)! where we u ed the binomial theorem and the ub titution j = k 1. Al o, we have n E(X(X 1)) = k=0 k(k 1) 2
n! pk (1 p)nk k!(n k)! (n 2)! pk2 (1 p)nk (k 2)!(n k)! (n 2)! pj (1 p)n n =n(n 1)p k=2 n2 =n(n 1)p2 j=0 2 =n(n 1)p (p + 1 p)n2 = n(n 1)p2
18 BINOMIAL AND MULTINOMIAL RANDOM VARIABLES 149
Thi implie E(X 2 ) = E(X(X 1)) + E(X) = n(n 1)p2 + np. The variation of X i t hen Var(X) = E(X 2 ) (E(X))2 = n(n 1)p2 + np n2 p2 = np(1 p) Example 18.7 The pr obability of hitting a target per hot i 0.2 and 10 hot are red independently. (a) What i the probability that the target i hit at lea t twice? (b) What i the expected number of ucce ful hot ? (c) How many hot mu t be red to make t he probability at lea t 0.99 that the target will be hit? Solution. Let X be the number of hot which hit the target. Then, X ha a binomial di tribution with n = 10 and p = 0.2. (a) The event that the target i hit at lea t twice i {X 2} . So, P (X 2) =1 P (X < 2) = 1 P (0) P (1) =1 C(10, 0)(0.2)0 (0.8)10 C(10, 1)(0 2)1 (0.8)9 0.6242 (b) E(X) = np = 10 (0.2) = 2. (c) Suppose that n shots are need ed to make the probability at least 0.99 that the target will be hit. Let A deno te that the target is hit. Then, Ac means that the target is not hit. We have, P (A) = 1 P (Ac ) = 1 (0.8)n 0.99 Solving the inequality, we nd that n number of hot i 21 ln (0.01) ln (0.8) 20.6. So, the required Example 18.8 Let X be a binomial random variable with parameter (12, 0.5). Find the variance and the tandard deviation of X. Solution. We have n = 12 and p = 0.5. Thu , Var(X) = np(1 p) = 6(1 0.5) = 3. The tandard deviation i SD(X) = 3
Example 18.9 An exam con i t of 25 multiple choice que tion in which there are ve choice for each que tion. Suppo e that you randomly pick an an wer for each que tion. Let X denote the total number of correctly an wered que tion . Write a n expre ion that repre ent each of the following probabilitie . (a) The probab ility that you get exactly 16, or 17, or 18 of the que tion correct. (b) The pr obability that you get at lea t one of the que tion correct. Solution. (a) P (X = 16 or X = 17 or X = 18) = C(25, 16)(0.2)16 (0.8)9 +C(25, 17)(0.2)17 (0.8)8 + C(25, 18)(0.2)18 (0.8)7 (b) P (X 1) = 1 P (X = 0) = 1 C(25, 0)(0.8)25 A u eful f act about the binomial di tribution i a recur ion for calculating the probabili ty ma function. Theorem 18.1 Let X be a binomial random variable with paramete r (n, p). Then for k = 1, 2, 3, , n p nk+1 p(k 1) p(k) = 1p k Proof. We have C(n k)pk (1 p)nk p(k) = p(k 1) C(n, k 1)pk1 (1 p)nk+1 n! pk (1 p)nk k!(nk)! = n )nk+1 (k1)!(nk+1)! = (n k + 1)p p nk+1 = k(1 p) 1p k The following theorem details how the binomial pmf rst increases and then decreas es. Theorem 18.2 Let X be a binomial random variable with parameters (n, p). As k goes from 0 to n, p(k) rst increases monotonically and then decreases monotonic ally reaching its largest value when k is the largest integer such that k (n + 1 )p.
18 BINOMIAL AND MULTINOMIAL RANDOM VARIABLES Proof. From the previous theorem we have p(k) p nk+1 (n + 1)p k = =1+ . p(k 1) 1p k k(1 p) 151 Accordingly, p(k) > p(k 1) when k < (n + 1)p and p(k) < p(k 1) when k > (n + 1)p . Now, if (n + 1)p = m is an integer then p(m) = p(m 1). If not, then by letting k = [(n + 1)p] = greatest integer less than or equal to (n + 1)p we nd that p re aches its maximum at k We illustrate the previous theorem by a histogram. Binomi al Random Variable Histogram The histogram of a binomial random variable is cons tructed by putting the r values on the horizontal axis and P (r) values on the v ertical axis. The width of the bar is 1 and its height is P (r). The bars are ce ntered at the r values. Example 18.10 Construct the binomial distribution for th e total number of heads in four ips of a balanced coin. Make a histogram. Solutio n. The binomial distribution is given by the following table 0 r 1 p(r) 16 1 4 16 2 6 16 3 4 16 4 1 16 The corresponding histogram is shown in Figure 18.1 Figure 18.1
152 DISCRETE RANDOM VARIABLES Multinomial Distribution A multinomial experiment is a statistical experiment th at has the following properties: The experiment consists of n repeated trials. E ach trial has two or more outcomes. On any given trial, the probability that a p articular outcome will occur is constant. The trials are independent; that is, t he outcome on one trial does not a ect the outcome on other trials. Example 18.11 Show that the experiment of rolling two dice three times is a mult inomial experiment.
Solution. The experiment consists of repeated trials. We toss the two dice three times. Each trial can result in a discrete number of outcomes 2 through 12. The probability of any outcome is constant; it does not change from one toss to the next. The trials are independent; that is, getting a particular outcome on one t rial does not a ect the outcome on other trials A multinomial distribution is the probability distribution of the outcomes from a multinomial experiment. Its prob ability mass function is given by the following multinomial formula: Suppose a m ultinomial experiment consists of n trials, and each trial can result in any of k possible outcomes: E1 , E2 , , Ek . Suppose, further, that each possible outco me can occur with probabilities p1 , p2 , , pk . Then, the probability P that E1 occurs n1 times, E2 occurs n2 times, , and Ek occurs nk times is n! n1 !n2 ! n ! pn1 pn2 pnk 1 2 k p(n1 , n2 , , nk , n, p1 , p2 , , pk ) =
18 BINOMIAL AND MULTINOMIAL RANDOM VARIABLES where n = n1 + n2 + + nk . Note tha t 153 (n1 , n2 , , nk ) n1 + n2 + nk = n n! pn1 pn2 pnk = (p1 + p2 + + pk )n = 1 k n1 !n2 ! nk ! 1 2 The examples below illustrate how to use the multinomial formula to compute the probability of an outcome from a multinomial experiment. Example 18.12 A factory makes three di erent kinds of bolts: Bolt A, Bolt B and Bolt C. The factory produ ces millions of each bolt every year, but makes twice as many of Bolt B as it do es Bolt A. The number of Bolt C made is twice the total of Bolts A and B combine d. Four bolts made by the factory are randomly chosen from all the bolts produce d by the factory in a given year. What is the probability that the sample will c ontain two of Bolt B and two of Bolt C? Solution. Because of the proportions in which the bolts are produced, a randomly 1 selected bolt will have a 9 chance of being of type A, a 2 chance of being of 9 2 type B, and a 3 chance of being of type C. A random selection of size n from the production of bolts will have a mu ltinomial distribution with parameters n, pA = 1 , pB = 2 , and pC = 2 , with pr obability mass function 9 9 3 P (NA = nA , NB = nB , Nc = nC ) = n! nA !nB !nC ! 1 9 nA 2 9 nB 2 3 nC Letting n = 4, nA = 0, nB = 2, nC = 2 we ! 1 9 0 2 9 2 2 3 2 = 32 243 Example 18.13 Suppose a card is drawn randomly from an ordinary deck of playing cards, and then put back in the deck. This exercise is repeated ve times. What is the probability of drawing 1 spade, 1 heart, 1 diamond, and 2 clubs? nd P (NA = 0, NB = 2, Nc = 2) = 4! 0!2!2
154 DISCRETE RANDOM VARIABLES Solution. To solve this problem, we apply the multinomial formula. We know the f ollowing: The experiment consists of 5 trials, so n = 5. The 5 trials produce 1 spade, 1 heart, 1 diamond, and 2 clubs; so n1 = 1, n2 = 1, n3 = 1, and n4 = 2. O n any particular trial, the probability of drawing a spade, heart, diamond, or c lub is 0.25, 0.25, 0.25, and 0.25, respectively. Thus, p1 = 0.25, p2 = 0.25, p3 = 0.25, and p4 = 0.25. We plug these inputs into the multinomial formula, as sho wn below: p(1, 1, 1, 2, 5, 0.25, 0.25, 0.25, 0.25) = 5! (0.25)1 (0.25)1 (0.25)1 (0.25)2 = 0.05859 1!1!1!2!
18 BINOMIAL AND MULTINOMIAL RANDOM VARIABLES 155 Practice Problems Problem 18.1 You are a telemarketer with a 10% chance of persuading a randomly s elected person to switch to your long distance company. You make 8 calls. What i s the probability that exactly one is sucessful? Problem 18.2 Tay Sachs disease is a rare but fatal disease of genetic origin. If a couple are both carriers of Tay Sachs disease, a child of theirs has probability 0.25 of being born with the disease. If such a couple has four children, what is the probability that 2 of the children have the disease? Problem 18.3 An internet service provider (IAP) o wns three servers. Each server has a 50% chance of being down, independently of the others. Let X be the number of servers which are down at a particular time. Find the probability mass function (PMF) of X. Problem 18.4 A hospital receives 1/5 of its u vaccine shipments from Company X and the remainder of its shipments from other companies. Each shipment contains a very large number of vaccine vial s. For Company Xs shipments, 10% of the vials are ine ective. For every other comp any, 2% of the vials are ine ective. The hospital tests 30 randomly selected vials from a shipment and nds that one vial is ine ective. What is the probability that this shipment came from Company X? Problem 18.5 A company establishes a fund of 120 from which it wants to pay an amount, C, to any of its 20 employees who achi eve a high performance level during the coming year. Each employee has a 2% chan ce of achieving a high performance level during the coming year, independent of any other employee. Determine the maximum value of C for which the probability i s less than 1% that the fund will be inadequate to cover all payments for high p erformance. Problem 18.6 (Trinomial Distribution) A large pool of adults earning their rst drivers license includes 50% low risk drivers, 30% moderate risk driver s, and 20% high risk drivers. Because these
156 DISCRETE RANDOM VARIABLES drivers have no prior driving record, an insurance company considers each driver to be randomly selected from the pool. This month, the insurance company writes 4 new policies for adults earning their rst drivers license. What is the probabi lity that these 4 will contain at least two more high risk drivers than low risk drivers?Hint: Use the multinomial theorem. Problem 18.7 A company prices its hu rricane insurance using the following assumptions: (i) In any calendar year, the re can be at most one hurricane. (ii) In any calendar year, the probability of a hurricane is 0.05 . (iii) The number of hurricanes in any calendar year is inde pendent of the number of hurricanes in any other calendar year. Using the compan ys assumptions, calculate the probability that there are fewer than 3 hurricanes in a 20 year period. Problem 18.8 1 The probability of winning a lottery is 300 . If you play the lottery 200 times, what is the probability that you win at lea st twice? Problem 18.9 Suppose an airline accepted 12 reservations for a commute r plane with 10 seats. They know that 7 reservations went to regular commuters w ho will show up for sure. The other 5 passengers will show up with a 50% chance, independently of each other. (a) Find the probability that the ight will be over booked. (b) Find the probability that there will be empty seats. Problem 18.10 S uppose that 3% of computer chips produced by a certain machine are defective. Th e chips are put into packages of 20 chips for distribution to retailers. What is the probability that a randomly selected package of chips will contain at least 2 defective chips? Problem 18.11 The probability that a particular machine brea ks down in any day is 0.20 and is independent of the breakdowns on any other day . The machine can break down only once per day. Calculate the probability that t he machine breaks down two or more times in ten days.
18 BINOMIAL AND MULTINOMIAL RANDOM VARIABLES 157 Problem 18.12 Coins K and L are weighted so the probabilities of heads are 0.3 a nd 0.1, respectively. Coin K is tossed 5 times and coin L is tossed 10 times. If all the tosses are independent, what is the probability that coin K will result in heads 3 times and coin L will result in heads 6 times? Problem 18.13 Custome rs at Freds Cafe win a 100 dollar prize if their cash register receipts show a st ar on each of the ve consecutive days Monday, , Friday in any one week. The cash r egister is programmed to print stars on a randomly selected 10% of the receipts. If Mark eats at Freds once each day for four consecutive weeks and the appearanc e of stars is an independent process, what is the standard deviation of X, where X is the number of dollars won by Mark in the four week period? Problem 18.14 I f X is the number of 6s that turn up when 72 ordinary dice are independently thrown , nd the expected value of X 2 . Problem 18.15 Suppose we have a bowl with 10 mar bles 2 red marbles, 3 green marbles, and 5 blue marbles. We randomly select 4 marbles from the bowl, with replacement. What is the probability of selecting 2 green marbles and 2 blue marbles? Problem 18.16 A tennis player nds that she wins against her best friend 70% of the time. They play 3 times in a particular mont h. Assuming independence of outcomes, what is the probability she wins at least 2 of the 3 matches? Problem 18.17 A package of 6 fuses are tested where the prob ability an individual fuse is defective is 0.05. (That is, 5% of all fuses manuf actured are defective). (a) What is the probability one fuse will be defective? (b) What is the probability at least one fuse will be defective? (c) What is the probability that more than one fuse will be defective, given at least one is de fective?
158 DISCRETE RANDOM VARIABLES Problem 18.18 In a promotion, a potato chip company inserts a coupon for a free bag in 10% of bags produced. Suppose that we buy 10 bags of chips, what is the p robability that we get at least 2 coupons? Problem 18.19 A tour operator has a b us that can accommodate 20 tourists. The operator knows that tourists may not sh ow up, so he sells 21 tickets. The probability that an individual tourist will n ot show up is 0.02, independent of all other tourists. Each ticket costs 50, and is non refundable if a tourist fails to show up. If a tourist shows up and a se at is not available, the tour operator has to pay 100 (ticket cost + 50 penalty) to the tourist. What is the expected revenue of the tour operator? Problem 18.2 0 An industrial salesperson has a salary structure as follows: S = 100 + 50Y 10Y 2 where Y is the number of items sold. Assuming sales are independent from one attempt to another, each with probability of success of p = 0.20. Give the expec ted salaries on days where the salesperson attempts n = 1, 2, and 3 sales.
19 POISSON RANDOM VARIABLE 159 19 Poisson Random Variable A random variable X is said to be a Poisson random variable with parameter > 0 i f its probabi ity mass function has the form k , k = 0, 1, 2, k! where indicates the average number of successes per unit time or space. Note that k p(k) = e = e 1. k! k=0 k=0 p(k) = e The Poisson random variab e is most common y used to mode the number of random occurrences of some phenomenon in a speci ed unit of space or time. For examp e, the number of phone ca s received by a te ephone operator i n a 10 minute period or the number of typos per page made by a secretary. Examp e 19.1 The number of fa se re a arms in a suburb of Houston averages 2.1 per day. Assuming that a Poisson distribution is appropriate, what is the probabi ity th at 4 fa se a arms wi occur on a given day? So ution. The probabi ity that 4 fa se a arms wi occur on a given day is given by P (X = 4) = e2.1 (2.1)4 0.0992 4 ! Examp e 19.2 Peop e enter a store, on average one every two minutes. (a) What is the probabi ity that no peop e enter between 12:00 and 12:05? (b) Find the prob abi ity that at east 4 peop e enter during [12:00,12:05]. So ution. (a) Let X b e the number of peop e that enter between 12:00 and 12:05. We mode X as a Poiss on random variab e with parameter , the average number of peop e that arrive in t he 5 minute interva . But, if 1 person arrives every 2 minutes, on average (so 1 /2 a person per minute), then in 5 minutes an average of 2.5 peop e wi arrive. Thus, = 2.5. Now, P (X = 0) = e2.5 2.50 = e2.5 . 0!
160 (b) DISCRETE RANDOM VARIABLES
P (X 4) =1 P (X 3) = 1 P (X = 0) P (X = 1) P (X = 2) P (X = 3) 2.51 2.52 2.53 50 e2.5 e2.5 e2.5 =1 e2.5 0! 1! 2! 3! Examp e 19.3 A ife insurance sa esman se n the average 3 ife insurance po icies per week. Use Poisson distribution to ca cu ate the probabi ity that in a given week he wi se (a) some po icies (b) 2 or more po icies but ess than 5 po icies. (c) Assuming that there are 5 worki ng days per week, what is the probabi ity that in a given day he wi se one p o icy? So ution. (a) Let X be the number of po icies so d in a week. Then P (X 1 ) = 1 P (X = 0) = 1 (b) We have P (2 X < 5) =P (X = 2) + P (X = 3) + P (X = 4) e3 32 e3 33 e3 34 = + + 0.61611 2! 3! 4! (c) Let X be the number of po icies so d pe r day. Then = P (X = 1) = e0.6 (0.6) 0.32929 1! 3 5 e3 30 0.95021 0! = 0.6. Thus, Examp e 19.4 A bio ogist suspects that the number of viruses in a sma vo ume o f b ood has a Poisson distribution. She has observed the proportion of such b oo d samp es that have no viruses is 0.01. Obtain the exact Poisson distribution sh e needs to use for her research. Make sure you identify the random variab e rst.
19 POISSON RANDOM VARIABLE 161 So ution. Let X be the number of viruses present. Then X is a Poisson distributi on with pmf P (X = k) = k e , k = 0, 1, 2, k! Since P (X = 0) = 0.01 we can write e = 0.01. Thus, = 4.605 and the exact Poisson distribution is P (X = k) = (4.605)k e4.605 , k = 0, 1, 2, k! Example 19.5 Assume that the number of defects in a certain type of magnetic tap e has a Poisson distribution and that this type of tape contains, on the average , 3 defects per 1000 feet. Find the probability that a roll of tape 1200 feet lo ng contains no defects. Solution. Let Y denote the number of defects on a 1200 f eet long roll of tape. Then Y is a Poisson random variable with parameter = 3 12 00 = 3.6. Thus, 1000 P (Y = 0) = e3.6 (3.6)0 = e3.6 0.0273 0! The expected va ue of X is found as fo ows. E(X) = = k k=1 e k k! e k1 (k 1)! e k k! k=1 = =e k=0 e =
162 DISCRETE RANDOM VARIABLES To nd the variance, we rst compute E(X 2 ).
E(X(X 1)) = = k(k 1) k=2 2 k=2 e k k! e k2 (k 2)! e k k! =2 k=0 =2 e e = 2 Thus, E(X 2 ) = E(X(X 1)) + E(X) = 2 + and Var(X) = E(X 2 ) (E(X))2 Examp e 19.6 In the inspection of a fabric produced in continuous ro s, there i s, on average, one imperfection per 10 square yards. Suppose that the number of imperfections is a random variab e having Poisson distribution. Let X denote the number of imperfections in a bo t of 50 square yards. Find the mean and the var iance of X. So ution. Since the fabric has 1 imperfection per 10 square yards, t he number of imperfections in a bo t of 50 yards is 5. Thus, X is a Poisson rand om variab e with parameter = 5. Hence, E(X) = Var(X) = = 5 Poisson Approximation to the Binomia Random Variab e. Next we show that a binomia random variab e w ith parameters n and p such that n is arge and p is sma can be approximated b y a Poisson distribution. Theorem 19.1 Let X be a binomia random variab e with parameters n and p. If n a nd p 0 so that np = = E(X) remains constant then X can be approximated by a Pois son distribution with parameter .
19 POISSON RANDOM VARIABLE Proof. First notice that for sma p << 1 we can writ e (1 p)n =en n (1p) =en(p 2 ) enp e We prove that np2 2 p2 163 k . k! This is true for k = 0 since P (X = 0) = (1 p)n e . Suppose k > 0 then P (X = k) e P (X = k) =C(n, k) n k 1 n nk n(n 1) (n k + 1) k = nk k! n n1 n k + 1 k n n n k! k 1 e 1 k 1 we have 1 n 1 n n 1 n 1 n k n k nj n j = 1 n 1 as n
364 365 100 0.2399.
In genera , Poisson distribution wi provide a good approximation to binomia p robabi ities when n 20 and p 0.05. When n 100 and np < 10, the approximation wi genera y be exce ent. Examp e 19.7 Let X denote the tota number of peop e w ith a birthday on Christmas day in a group of 100 peop e. Then X is a binomia r andom variab e with 1 parameters n = 100 and p = 365 . What is the probabi ity a t east one person in the group has a birthday on Christmas day? So ution. We ha ve P (X 1) = 1 P (X = 0) = 1 C(100, 0) 1 365 0
164 DISCRETE RANDOM VARIABLES 1 365 Using the Poisson approximation, with = 100 = 100 365 = 20 73 we nd
(20/73)0 ( 20 ) P (X 1) = 1 P (X = 0) = 1 e 73 0.2396 0! Examp e 19.8 Suppose we ro two dice 12 times and we et X be the number of times a doub e 6 appears. H ere n = 12 and p = 1/36, so = np = 1/3. For k = 0, 1, 2 compare P (X = k) found using the binomia distribution with the one found using Poisson approximation. So ution. Let PB (X = k) be the probabi ity using binomia distribution and P (X = k) be the probabi ity using Poisson distribution. Then PB (X = 0) = 1 1 36 1 12 0.7132 P (X = 0) = e 3 0.7165 1 1 PB (X = 1) = C(12, 1) 1 0.2445 36 36 1 1 P (X = 1) = e 3 0.2388 3 2 10 1 1 PB (X = 2) = C(12, 2) 0.0384 1 36 36 P (X = 2) = e 3 1 11 1 3 2 1 0.0398 2! Examp e 19.9 An archer shoots arrows at a circu ar target where the centra port ion of the target inside is ca ed the bu . The archer hits the bu with proba bi ity 1/32. Assume that the archer shoots 96 arrows at the target, and that a shoots are independent. (a) Find the probabi ity mass function of the number of bu s that the archer hits. (b) Give an approximation for the probabi ity of th e archer hitting no more than one bu .
19 POISSON RANDOM VARIABLE 165 So ution. (a) Let X denote the number of shoots that hit the bu . Then X is bin omia y distributed: P (X = k) = C(n, k)pk (1 p)nk , n = 96, p = 1 . 32 (b) Since n is arge, and p sma , we can use the Poisson approximation, with pa rameter = np = 3. Thus, P (X 1) = P (X = 0) + P (X = 1) e + e = 4e3 0.199 We this section by estab ishing a recursion formu a for computing p(k). Theorem 19. 2 If X is a Poisson random variab e with parameter , then p(k + 1) = Proof. We ha ve p(k + 1) e (k+1)! = k p(k) e k! = k+1 k+1 p(k). k+1
166 DISCRETE RANDOM VARIABLES Practice Prob ems Prob em 19.1 Suppose the average number of car accidents on the highway in one d ay is 4. What is the probabi ity of no car accident in one day? What is the prob abi ity of 1 car accident in two days? Prob em 19.2 Suppose the average number o f ca s received by an operator in one minute is 2. What is the probabi ity of r eceiving 10 ca s in 5 minutes? Prob em 19.3 In one particu ar autobiography of a professiona ath ete, there are an average of 15 spe ing errors per page. If the Poisson distribution is used to mode the probabi ity distribution of the nu mber of errors per page, then the random variab e X, the number of errors per pa ge, has a Poisson distribution with = 15. What is the probabi ity of having no e rrors on a page? Prob em 19.4 Suppose that the number of accidents occurring on a highway each day is a Poisson random variab e with parameter = 3. (a) Find the probabi ity that 3 or more accidents occur today. (b) Find the probabi ity that no accidents occur today. Prob em 19.5 At a checkout counter customers arrive a t an average of 2 per minute. Find the probabi ity that (a) at most 4 wi arriv e at any given minute (b) at east 3 wi arrive during an interva of 2 minutes (c) 5 wi arrive in an interva of 3 minutes. Prob em 19.6 Suppose the number of hurricanes in a given year can be mode ed by a random variab e having Poisson distribution with standard deviation = 2. What i the probability that there ar e at lea t three hurricane ? Problem 19.7 A Geiger counter i monitoring the lea kage of alpha particle from a container of radioactive material. Over a long pe riod of time, an average of
19 POISSON RANDOM VARIABLE 167 50 particle per minute i mea ured. A ume the arrival of particle at the coun ter i modeled by a Poi on di tribution. (a) Compute the probability that at le a t one particle arrive in a particular one econd period. (b) Compute the prob ability that at lea t two particle arrive in a particular two econd period. Pr oblem 19.8 A type etter, on the average, make one error in every 500 word type et. A typical page contain 300 word . What i the probability that there will be no more than two error in ve page ? Problem 19.9 An ctu ry h s discovered th t policyholders re three times s likely to le two cl ims s to le four cl ims. If the number of cl ims led h s Poisson distribution, wh t is the v ri nce of t he number of cl ims led? Problem 19.10 A comp ny buys policy to insure its reve nue in the event of m jor snowstorms th t shut down business. The policy p ys no thing for the rst such snowstorm of the ye r nd $10,000 for e ch one there fter, until the end of the ye r. The number of m jor snowstorms per ye r th t shut do wn business is ssumed to h ve Poisson distribution with me n 1.5 . Wh t is th e expected mount p id to the comp ny under this policy during one ye r period ? Problem 19.11 A b seb ll te m h s scheduled its opening g me for April 1. If i t r ins on April 1, the g me is postponed nd will be pl yed on the next d y th t it does not r in. The te m purch ses insur nce g inst r in. The policy will p y 1000 for e ch d y, up to 2 d ys, th t the opening g me is postponed. The insu r nce comp ny determines th t the number of consecutive d ys of r in beginning o n April 1 is Poisson r ndom v ri ble with me n 0.6 . Wh t is the st nd rd devi tion of the mount the insur nce comp ny will h ve to p y?
168 DISCRETE RANDOM VARIABLES Problem 19.12 The ver ge number of trucks rriving on ny one d y t truck de pot in cert in city is known to be 12. Wh t is the prob bility th t on given d y fewer th n nine trucks will rrive t this depot? Problem 19.13 A cert in k ind of sheet met l h s, on the ver ge, ve defects per 10 squ re feet. If we ssu me Poisson distribution, wh t is the prob bility th t 15 squ re feet sheet o f the met l will h ve t le st six defects? Problem 19.14 Let X be Poisson r n dom v ri ble with me n . If P (X = 1 X 1) = 0.8, what is the value of ? Prob em 19 .15 The number of de iveries arriving at an industria warehouse (during work ho urs) has a Poisson distribution with a mean of 2.5 per hour. (a) What is the pro babi ity an hour goes by with no more than one de ivery? (b) Give the mean and s tandard deviation of the number of de iveries during 8 hour work days. Prob em 1 9.16 Imperfections in ro s of fabric fo ow a Poisson process with a mean of 64 per 10000 feet. Give the approximate probabi ity a ro wi have at east 75 i mperfections. Prob em 19.17 At a certain retai er, purchases of ottery tickets in the na 10 minutes of sa e before a draw fo ow a Poisson distribution with = 15 if the top prize is ess than $10,000,000 and fo ow a Poisson distribution w ith = 10 if the top prize is at east $10,000,000. Lottery records indicate that the top prize is $10,000,000 or more in 30% of draws. (a) Find the probabi ity that exact y 15 tickets are so d in the na 10 minutes before a ottery draw. (b) If exact y 15 tickets are so d in the na 10 minutes of sa e before the draw, wh at is the probabi ity that the top prize in that draw is $10,000,000 or more?
20 OTHER DISCRETE RANDOM VARIABLES 169 20 Other Discrete Random Variab es 20.1 Geometric Random Variab e A geometric random variab e with parameter p, 0 < p < 1 has a probabi ity mass f unction p(n) = p(1 p)n1 , n = 1, 2, . A geometric random variable models the numb er of successive Bernoulli trials that must be performed to obtain the rst success. For example, the number of ips of a fair coin until the rst head appears follows a geometric distribution. Example 20.1 1 If you roll a pair of fair dice, the pr obability of getting an 11 is 18 . If you roll the dice repeatedly, what is the probability that the rst 11 occurs on the 8th roll? Solution. Let X be the number of rolls on which the rst 11 occurs. Then X is a 1 geometric random variable wit h parameter p = 18 . Thus, P (X = 8) = 1 18 1 1 18 7 = 0.0372
To nd the expected value and variance of a geometric random variable we 1 proceed as follows. First we recall that for 0 < x < 1 we have xn = 1x . n=0 Di erentiatin g this twice we nd n=1 nxn1 = (1 x)2 and n=1 n(n 1)xn2 = 2(1 x)3 . Letting x = 1 the last two equations we nd n=1 n(1 p)n1 = p2 and n=1 n(n 1)(1 p)n2 = 2p3 . We next apply these equalities in nding E(X) and E(X 2 ). Indeed, we have E(X) = n=1 n(1 p)n1 p n(1 p)n1 n=1 2 =p =p p = p1
170 and DISCRETE RANDOM VARIABLES E(X(X 1)) = n=1 n(n 1)(1 p)n1 p =p(1 p) n=1 n(n 1)(1 p)n2
=p(1 p) (2p3 ) = 2p2 (1 p) so that E(X 2 ) = (2 p)p2 . The variance is then give y Var(X) = E(X 2 ) (E(X))2 = (2 p)p2 p2 = Note that 1p . p2 p(1 p)n1 = n=1 p = 1. 1 (1 p) Also, observe that for k = 1, 2, we have P (X k) = n=k p(1p) n1 = p(1p) k1 n=0 (1p)n = p(1 p)k1 = (1p)k1 1 (1 p) and P (X k) = 1 P (X k + 1) = 1 (1 p)k . From this, one can nd the cdf given by (X) = P (X x) = 0 x<1 1 (1 p)k k x < k + 1, k = 1, 2,
We can use F (x) for computing probabilities such as P (j X k), which is the pro bability that it will take from j attempts to k attempts to succeed. This value is given by P (j X k) =P (X k) P (X j 1) = (1 (1 p)k ) (1 (1 p)j1 ) k Remark 20.1 The fact that P (j X k) = P (X k) P (X j 1) follows from Proposit on 21.9 of Section 21.
20 OTHER DISCRETE RANDOM VARIABLES 171 Example 20.2 Computer chips are tested one at a time until a good chip is found. Let X denote the number of chips that need to be tested in order to nd a good on e. Given that P (X > 3) = 1/2, what is E(X)? Solution. X has geometric distribut ion, so P (X > 3) = P (X 4) = (1 p)3 . Setting 1 this equa to 1/2 and so ving f or p gives p = 1 2 3 . Therefore, E(X) = 1 1 = 1 p 1 2 3 Examp e 20.3 From past experience it is known that 3% of accounts in a arge acc ounting popu ation are in error. (a) What is the probabi ity that 5 accounts are audited before an account in error is found? (b) What is the probabi ity that t he rst account in error occurs in the rst ve accounts audited? So ution. (a) Let X be the number of inspections to obtain rst account in error. Then X is a geometri c random variab e with p = 0.03. Thus P (X = 5) = (0.03)(0.97)4 . (b) P (X 5) = 1 P (X 6) = 1 (0.97)5 0.141 Examp e 20.4 Suppose you keep ro ing four dice simu taneous y unti at east one of them shows a 6. What is the expected number of r ounds (each round consisting of a simu taneous ro of four dice) you have to p a y? So ution. If success of a round is de ned as at east one 6 then X is simp y the 4 number of tria s needed to get the rst success, with p given by p = 1 5 6 1 0.5177 5. Thus, X has geometric distribution, and so E(X) = p 1.9314
172 DISCRETE RANDOM VARIABLES Examp e 20.5 Assume that every time you attend your 2027 ecture there is a prob abi ity of 0.1 that your Professor wi not show up. Assume her arriva to any g iven ecture is independent of her arriva (or non arriva ) to any other ecture . What is the expected number of c asses you must attend unti you arrive to nd y our Professor absent? So ution. Let X be the number of c asses you must attend u nti you arrive to nd your Professor absent, then X has Geometric distribution wi th parameter p = 0.1. Thus P (X = n) = 0.1(1 p)n1 , n = 1, 2, and E(X) = 1 = 10 p
20 OTHER DISCRETE RANDOM VARIABLES 173 Practice Problems Problem 20.1 An urn contains 5 white, 4 black, and 1 red marble. Marbles are dra wn, with replacement, until a red one is found. If X is the random variable coun ting the number of trials until a red marble appears, then (a) What is the proba bility that the red marble appears on the rst trial? (b) What is the probability that the red marble appears on the second trial? (c) What is is the probability that the marble appears on the kth trial. Problem 20.2 The probability that a ma chine produces a defective item is 0.10. Each item is checked as it is produced. Assume that these are independent trials, nd the probability that at least 10 it ems must be checked to nd one that is defective. Problem 20.3 Suppose parts are o f two varieties: good (with probability 90/92) and slightly defective (with prob ability 2/92). Parts are produced one after the other. What is the probability t hat at least 4 good parts will be produced before the rst defective part? Problem 20.4 Assume that every time you drive your car, there is a 0.001 probability th at you will receive a speeding ticket, independent of all other times you have d riven. (a) What is the probability you will drive your car two or less times bef ore receiving your rst ticket? (b) What is the expected number of times you will drive your car until you receive your rst ticket? Problem 20.5 Three people study together. They dont like to ask questions in class, so they agree that they will ip a fair coin and if one of them has a di erent outcome than the other two, that person will ask the groups question in class. If all three match, they ip again un til someone is the questioner. What is the probability that (a) exactly three ro unds of ips are needed? (b) more than four rounds are needed?
174 DISCRETE RANDOM VARIABLES Problem 20.6 Suppose one die is rolled over and over until a Two is rolled. What is the probability that it takes from 3 to 6 rolls? Problem 20.7 You have a fai r die that you roll over and over until you get a 5 or a 6. (You stop rolling it once it shows a 5 or a 6.) Let X be the number of times you roll the die. (a) W hat is P (X = 3)? What is P (X = 50)? (b) What is E(X)? Problem 20.8 Fifteen per cent of houses in your area have nished basements. Your real estate agent starts showing you homes at random, one after the other. Let X be the number of homes w ith nished basements that you see, before the rst house that has no nished basement . (a) What is the probability distribution of X? (b) What is the probability dis tribution of Y = X + 1? Problem 20.9 Suppose that 3% of computer chips produced by a certain machine are defective. The chips are put into packages of 20 chips for distribution to retailers. (a) What is the probability that a randomly selec ted package of chips will contain at least 2 defective chips? (b) Suppose we con tinue to select packs of bulbs randomly from the production. What is the probabi lity that it will take fewer than ve packs to nd a pack with at least 2 defective chips? Problem 20.10 Show that the Geometric distribution with parameter p satis e s the equation P (X > i + j X > i) = P (X > j). This says that the Geometric dis tribution satis es the memoryless property Problem 20.11 As part of the underwriti ng process for insurance, each prospective policyholder is tested for high blood pressure. Let X represent the number of tests
20 OTHER DISCRETE RANDOM VARIABLES 175 completed when the rst person with high blood pressure is found. The expected val ue of X is 12.5. Calculate the probability that the sixth person tested is the rs t one with high blood pressure. Problem 20.12 An actress has a probability of ge tting o ered a job after a try out of p = 0.10. She plans to keep trying out for n ew jobs until she gets o ered. Assume outcomes of try outs are independent. (a) Ho w many try outs does she expect to have to take? (b) What is the probability she will need to attend more than 2 try outs? Problem 20.13 Which of the following sampling plans is binomial and which is geometric? Give P (X = x) = p(x) for eac h case. Assume outcomes of individual trials are independent with constant proba bility of success (e.g. Bernoulli trials). (a) A missile designer will launch mi ssiles until one successfully reaches its target. The underlying probability of success is 0.40. (b) A clinical trial enrolls 20 children with a rare disease. E ach child is given an experimental therapy, and the number of children showing m arked improvement is observed. The true underlying probability of success is 0.6 0.
20.2 Negative Binomial Random Variable Consider a statistical experiment where a success occurs with probability p and a failure occurs with probability q = 1 p. If the experiment is repeated inde nite ly and the trials are independent of each other, then the random variable X, the number of trials at which the rth success occurs, has a negative binomial distr ibution with parameters r and p. The probability mass function of X is p(n) = P (X = n) = C(n 1, r 1)pr (1 p)nr , where n = r, r + 1, (In order to have r succ es there must be at least r trials.) For the rth success to occur on the nth tri al, there must have been r 1 successes and n r failures among the rst n 1 trials. The number of ways of distributing r 1 successes among n 1 trials is C(n 1, r 1 ). But the probability of having r 1 successes and n r failures is pr1 (1 p)nr . T he probability of the rth success is p. Thus, the product of these three terms i s the probability that there are r successes and n r failures in the n trials, w ith the rth success occurring on the nth trial. Note that if r = 1 then X is a g eometric random variable with parameter p. The negative binomial distribution is sometimes de ned in terms of the random variable Y = number of failures before th e rth success. This formulation is statistically equivalent to the one given abo ve in terms of X = number of trials at which the rth success occurs, since Y = X r. The alternative form of the negative binomial distribution is P (Y = y) = C( r + y 1, y)pr (1 p)y , y = 0, 1, 2, . In this form, the negative binomial distri bution is used when the number of successes is xed and we are interested in the n umber of failures before reaching the xed number of successes. Note that C(r + y 1, y) = C(r + y 1, r 1). The negative binomial distribution gets its name from t he relationship C(r + y 1, y) = (1)y (r)(r 1) (r y + 1) = (1)y C(r, y) y!
20 OTHER DISCRETE RANDOM VARIABLES 177 which is the de ning equation for binomial coe cient with negative integers. Now, re calling the binomial series expansion (1 t) r = k=0 (1)k C(r, k)tk C(r + k 1, k)tk , k=0 = Thus, 1<t<1 P (Y = y) = y=0 C(r + y 1, y)pr (1 p)y y=0 r =p C(r + y 1, y)(1 p)y =1 =p p r y=0 r This shows that p(n) is indeed a probability mass function. Example 20.6 A resea rch scientist is inoculating rabbits, one at a time, with a disease until he nds two rabbits which develop the disease. If the probability of contracting the dis ease 1 , what is the probability that eight rabbits are needed? 6 Solution. Let X be the number of rabbits needed until the rst rabbit to contract the disease. T hen X follows a negative binomial distribution with r = 2, n = 6, 1 and p = 6 . Thus, P (8 rabbits are needed) = C(2 1 + 6, 6) 1 6 2 5 6 6 0.0651 Example 20.7 Suppose we are at a ri e range with an old gun that mis res 5 out of 6 times. De ne success as the event the gun res and let Y be the number of failures bef ore the third success. What is the probability that there are 10 failures before the third success?
178 DISCRETE RANDOM VARIABLES Solution. The probability that there are 10 failures before the third success is given by P (Y = 10) = C(12, 10) 1 6 3 5 6 10 0.0493 Example 20.8 A comedian tells jokes that are only funny 25% of the time. What is the probability that he tells his tenth funny joke in his fortieth attempt? Sol ution. A success is when a joke is funny. Let X number of attempts at which the tenth joke occurs. Then X is a negative binomial random variable with parameters r = 10 and p = 0.25. Thus, P (X = 40) = C(40 1, 10 1)(0.25)10 (0.75)30 0.036091 1 The expected value of Y is E(Y ) = y=0
yC(r + y 1, y)pr (1 p)y (r + y 1)! r p (1 p)y (y 1)!(r 1)! r(1 p) C(r + y 1 )pr+1 (1 p)y1 p = y=1 = y=1 r(1 p) = p = It follows that r(1 p) p C((r + 1) + z 1, z)pr+1 (1 p)z z=0 r E(X) = E(Y + r) = E(Y ) + r = . p
20 OTHER DISCRETE RANDOM VARIABLES Similarly, 179 E(Y 2 ) = y=0 y 2 C(r + y + 1, y)pr (1 p)y r(1 p) p r(1 p) p = = y y=1 (r + y 1)! r+1 p (1 p)y1 (y 1)!r! (z + 1)C((r + 1) + z 1, z)pr+1 (1 p)z z=0 r(1 p) (E(Z) + 1) meters r + 1 and p. ial random variable . E(Y ) = p2 p2 The r, Var(X) = Var(Y ) = p where Z is the negative binomial random variable with para Using the formula for the expected value of a negative binom gives that (r + 1)(1 p) E(Z) = p Thus, r2 (1 p)2 r(1 p) 2 + variance of Y is Var(Y ) = E(Y 2 ) [E(Y )]2 = Since X = Y + = r(1 p) . p2 r(1 p) . p2
Example 20.9 Suppose we are at a ri e range with an old gun that mis res 5 out of 6 times. De ne success as the event the gun res and let X be the number of failures bef ore the third success. Then X is a negative binomial random 1 variable with para mters (3, 6 ). Find E(X) and Var(X). Solution. The expected value of X is E(X) = r(1 p) = 15 p
180 and the variance is Var(X) = DISCRETE RANDOM VARIABLES r(1 p) = 90 p2
20 OTHER DISCRETE RANDOM VARIABLES 181 Practice Problems Problem 20.14 A phenomenal major league baseball player has a batting average of 0.400. Beginning with his next at bat, the random variable X, whose value refer s to the number of the at bat (walks, sacri ce ies and certain types of outs are no t considered at bats) when his rth hit occurs, has a negative binomial distribut ion with parameters r and p = 0.400. What is the probability that this hitters se cond hit comes on the fourth at bat? Problem 20.15 A geological study indicates that an exploratory oil well drilled in a particular region should strike oil wi th p = 0.2. Find P(the 3rd oil strike comes on the 5th well drilled). Problem 20 .16 We randomly draw a card with replacement from a 52 card deck and record its face value and then put it back. Our goal is to successfully choose three aces. Let X be the number of trials we need to perform in order to achieve our goal. ( a) What is the distribution of X? (b) What is the probability that X = 39? Probl em 20.17 Find the expected value and the variance of the number of times one mus t throw a die until the outcome 1 has occurred 4 times. Problem 20.18 Find the p robability that a man ipping a coin gets the fourth head on the ninth ip. Problem 20.19 The probability that a driver passes the written test for a drivers license is 0.75. What is the probability that a person will fail the test on the rst try and pass the test on the second try? Problem 20.20 A company takes out an insur ance policy to cover accidents that occur at its manufacturing plant. The probab ility that one or more accidents will occur
182 DISCRETE RANDOM VARIABLES during any given month is 3 . 5 The number of accidents that occur in any given month is independent of the number of accidents that occur in all other months. Calculate the probability that there will be 4 (or more) months, not necessarily consecutive, with no accidents before 4 months, not necessarily consecutive, wi th at least one accident. Problem 20.21 Waiters are extremely distracted today a nd have a 0.6 chance of making a mistake with your co ee order, giving you decaf e ven though you order ca einated. Find the probability that your order of the secon d deca einated co ee occurs on the seventh order of regular co ees. Problem 20.22 Supp ose that 3% of computer chips produced by a certain machine are defective. The c hips are put into packages of 20 chips for distribution to retailers. (a) What i s the probability that a randomly selected package of chips will contain at leas t 2 defective chips? (b) What is the probability that the tenth pack selected is the third to contain at least two defective chips? Problem 20.23 Assume that ev ery time you attend your 2027 lecture there is a probability of 0.1 that your Pr ofessor will not show up. Assume her arrival to any given lecture is independent of her arrival (or non arrival) to any other lecture. What is the expected numb er of classes you must attend until the second time you arrive to nd your Profess or absent? Problem 20.24 If the probability is 0.40 that a child exposed to a ce rtain contagious disease will catch it, what is the probability that the tenth c hild exposed to the disease will be the third to catch it? Problem 20.25 In roll ing a fair die repeatedly (and independently on successive rolls), nd the probabi lity of getting the third 1 on the t th roll.
20 OTHER DISCRETE RANDOM VARIABLES 183 20.3 Hypergeometric Random Variable Suppose we have a population of N objects which are divided into two types: Type A and Type B. There are n objects of Type A and N n objects of Type B. For exam ple, a standard deck of 52 playing cards can be divided in many ways. Type A cou ld be Hearts and Type B could be All Others. Then there are 13 Hearts and 39 others in this population of 52 cards. Suppose a random sample of size r is taken (with out replacement) from the entire population of N objects. The Hypergeometric ran dom variable X counts the total number of objects of Type A in the sample. If r n then there could be at most r objects of Type A in the sample. If r > n, then there can be at most n objects of Type A in the sample. Thus, the value min{r, n } is the maximum possible number of objects of Type A in the sample. On the othe r hand, if r N n, then all objects chosen may be of Type B. But if r > N n, then there must be at least r (N n) objects of Type A chosen. Thus, the value max{0, r (N n)} is the least possible number of objects of Type A in the sample. What is the probability of having exactly k objects of Type A in the sample, where ma x{0, r (N n)} k min{r, n}? This is a type of problem that we have done before: I n a group of N people there are n men (and the rest women). If we appoint a comm ittee of r persons from this group at random, what is the probability there are exactly k men on it? There are C(N, r) rsubsets of the group. There are C(n, k) ks ubsets of the men and C(N r, r k) (r k) subsets of the women. Thus the probabili ty of getting exactly k men on the committee is C(n, k)C(N n, r k) , k = 0, 1, , r, r < min{n, N n} p(k) = P (X = k) = C(N, r) This is the probability mass funct ion of X. Note that r k=0 C(n, k)C(N n, r k) =1 C(N, r) The proof of this result follows from Vendermondes identity Theorem 20.1 r C(n + m, r) = k=0 C(n, k)C(m, r k)
184 DISCRETE RANDOM VARIABLES Proof. Suppose a committee consists of n men and m women. In how many ways can a subcommittee of r members be formed? The answer is C(n + m, r). But on the othe r hand, the answer is the sum over all possible values of k, of the number of su bcommittees consisting of k men and r k women Example 20.10 Suppose a sack conta ins 70 red beads and 30 green ones. If we draw out 20 without replacement, what is the probability of getting exactly 14 red ones? Solution. If X is the number of red beads, then X is a hypergeometric random variable with parameters N = 100 , r = 20, n = 70. Thus, P (X = 14) = C(70, 14)C(30, 6) 0.21 C(100, 20) Example 20.11 13 Democrats, 12 Republicans and 8 Independents are sitting in a r oom. 8 of these people will be selected to serve on a special committee. What is the probability that exactly 5 of the committee members will be Democrats? Solu tion. Let X be the number of democrats in the committee. Then X is hypergeometri c random variable with parameters N = 33, r = 8, n = 13. Thus, P (X = 5) = C(13, 5)C(20, 3) 0.00556 C(33, 8) Next, we nd the expected value of a hypergeometric random variable with parameter s N, n, r. Let k be a positive integer. Then r E(X ) = i=0 r k ik P (X = i) ik i=0 = Using the identities C(n, i)C(N n, r i) C(N, r) iC(n, i) = nC(n 1, i 1) and rC(N, r) = N C(N 1, r 1)
20 OTHER DISCRETE RANDOM VARIABLES we obtain that E(X k ) = = nr N r 185 ik1 i=1 r1 C(n 1, i 1)C(N n, r i) C(N 1, r 1)
C(n 1, j)C((N 1) (n 1), (r 1) j) nr (j + 1)k1 N j=0 C(N 1, r 1) nr = E[(Y N where Y is a hypergeometric random variable with paramters N 1, n 1, and r 1. By taking k = 1 we nd E(X) = Now, by setting k = 2 we nd E(X 2 ) = Hence, V ar(X) = E(X 2 ) [E(X)]2 = nr (n 1)(r 1) nr +1 . N N 1 N nr nr E(Y + 1) = N N (n 1)(r +1 . N 1 nr . N Example 20.12 The United States senate has 100 members. Suppose there are 54 Rep ublicans and 46 democrats. A committee of 15 senators is selected at random. (a) What is the probability that there will be 9 Republicans and 6 Democrats on the committee? (b) What is the expected number of Republicans on this committee? (c ) What is the variance of the number of Republicans on this committee? Solution. Let X be the number of republicans of the committee of 15 selected at random. T hen X is a hypergeometric random variable with N = 100, r = 54, and n = 15.. (a) P (X = 9) = C(54,9)C(46,6) . C(100,15) 54 (b) E(X) = nr = 15 100 = 8.1 N r r (c) Var(X) = n N NN N n = 3.199 N 1
186 DISCRETE RANDOM VARIABLES Example 20.13 A lot of 15 semiconductor chips contains 6 defective chips and 9 g ood chips. Five chips are randomly selected without replacement. (a) What is the probability that there are 2 defective and 3 good chips in the sample? (b) What is the probability that there are at least 3 good chips in the sample? (c) What is the expected number of defective chips in the sample? Solution. (a) Let X be the number of defective chips in the sample. Then, X has a hypergeometric distribution with r = 6, N = 15, n = 5. The desired probability is 420 C(6, 2)C(9, 3) = P (X = 2) = C(15, 5) 1001 (b) Note that the event that there are at least 3 good chips in the sample is equivalent to the event that th ere are at most 2 defective chips in the sample, i.e. {X 2}. So, we have P (X 2) =P (X = 0) + P (X = 1) + P (X = 2) C(6, 0)C(9, 5) C(6, 1)C(9, 4) C(6, 2)C(9, 3) + + = C(15, 5) C(15, 5) C(15, 5) 714 = 1001 (c) E(X) = rn N =5 6 15 =2
20 OTHER DISCRETE RANDOM VARIABLES 187 Practice Problems Problem 20.26 Suppose we randomly select 5 cards without replacement from an ord inary deck of playing cards. What is the probability of getting exactly 2 red ca rds (i.e., hearts or diamonds)? Problem 20.27 There are 25 socks in a box, 15 re d and 10 blue. Pick 7 without replacement. What is the probability of picking ex actly 3 red socks? Problem 20.28 A player in the California lottery chooses 6 nu mbers from 53 and the lottery o cials later choose 6 numbers at random. Let X equa l the number of matches. Find the probability distribution function. Problem 20. 29 Compute the probability of obtaining 3 defectives in a sample of size 10 with out replacement from a box of 20 components containing 4 defectives. Problem 20. 30 A bag contains 10 $50 bills and 190 $1 bills, and 10 bills are drawn without replacement, what is the probability that 2 of the bills drawn are $50s? Problem 20.31 A package of 8 AA batteries contains 2 batteries that are defective. A st udent randomly selects four batteries and replaces the batteries in his calculat or. (a) What is the probability that all four batteries work? (b) What are the m ean and variance for the number of batteries that work? Problem 20.32 Suppose a company eet of 20 cars contains 7 cars that do not meet government exhaust emissi ons standards and are therefore releasing excessive pollution. Moreover, suppose that a tra c policeman randomly inspects 5 cars. What is the probability of no mo re than 2 polluting cars being selected? Problem 20.33 There are 123,850 pickup trucks in Austin. Of these, 2,477 are stolen. Suppose that 100 randomly chosen p ickups are checked by the police. What is the probability that exactly 3 of the 100 chosen pickups are stolen? Give the expression without nding the exact numeri c value.
188 DISCRETE RANDOM VARIABLES Problem 20.34 Consider an urn with 7 red balls and 3 blue balls. Suppose we draw 4 balls without replacement and let X be the total number of red balls we get. Compute P (X 1). Problem 20.35 As part of an air pollution survey, an inspector decided to examine the exhaust of six of a companys 24 trucks. If four of the com panys trucks emit excessive amounts of pollutants, what is the probability that n one of them will be included in the inspectors sample? Problem 20.36 A fair die i s tossed until a 2 is obtained. If X is the number of trials required to obtain the rst 2, what is the smallest value of x for which P (X x) 1 ? 2 Prob em 20.37 A box contains 10 white and 15 b ack marb es. Let X denote the number of white m arb es in a se ection of 10 marb es se ected at random and without rep acement. Find Var(X) . E(X) Prob em 20.38 Among the 48 app icants for a job, 30 have co ege degrees. 10 of the app icants are random y chosen for interviews. Let X be t he number of app icants among these ten who have co ege degrees. Find P (X 8).
21 PROPERTIES OF THE CUMULATIVE DISTRIBUTION FUNCTION189
21 Properties of the Cumu ative Distribution Function Random variab es are c assi ed into three types rather than two: Continuous, discr ete, and mixed. A random variab e wi be ca ed a continuous type random variab e if its cumu ative distribution function is continuous. As a matter of fact, i ts cumu ative distribution function is a continuous nondecreasing function. Thus , its cumu ative distribution graph has no jumps. A thorough discussion about th is type of random variab es starts in the next section. On the other extreme are the discrete type random variab es, which are a about jumps as you wi notic e ater on in this section. Indeed, the graph of the cumu ative distribution fun ction is a step function. A random variab e whose cumu ative distribution functi on is part y discrete and part y continuous is ca ed a mixed random variab e. I n this section, we wi discuss properties of the cumu ative distribution functi on that are va id to a three types. Reca from Section 14 that if X is a rand om variab e (discrete or continuous) then the cumu ative distribution function ( abbreviated c.d.f) is the function F (t) = P (X t). In this section we discuss s ome of the properties of c.d.f and its app ications. First, we prove that probab i ity is a continuous set function. In order to do that, we need the fo owing d e nitions. A sequence of sets {En } is said to be increasing if n=1 E1 E2 En hereas it is said to be a decreasing sequence if E1 E2 En En+1 If {En easing sequence of events we de ne a new event n=1 n lim En = En . n=1 For a decreasing sequence we de ne n lim En = En . n=1
190 DISCRETE RANDOM VARIABLES Proposition 21.1 If {En }n1 is either an increasing or decreasing sequence of eve nts then (a) n im P (En ) = P ( im En ) n that is P ( En ) = im P (En ) n=1 n for increasing sequence and (b) P ( En ) = n for decreasing sequence. Proof. (a) Suppose rst that En En+1 for a n1 , n > 1
Figure 21.1 These events are shown in the Venn diagram of Figure 21.1. Note that for n > 1, Fn consists of those outcomes in En that are not in any of the ear i er Ei , i < n. C ear y, for i = j we have Fi Fj = . A so, Fn = En n=1 n=1
im P (En ) n=1
n 1. De ne the events F1 =E1 c Fn =En E
21 PROPERTIES OF THE CUMULATIVE DISTRIBUTION FUNCTION191 and for n 1 we have n Fi = n Ei . From these properties we have i=1 i=1 P ( im En ) =P ( En ) n=1 n =P ( Fn ) n=1 = n=1 P (Fn ) n = im n P (Fn ) i=1 = im P (n Fi ) i=1 n = im P (n Ei ) i=1 n = im P (En ) n (b) an c c 1 c
Now suppose that {En }n1 is a decreasing sequence of events. Then c {En }n1 is increasing sequence of events. Hence, from part (a) we have P ( En ) = im P (En ) n=1 n c By De Morgans Law we have En = ( En )c . Thus, P (( En )c ) = im P (En ). n=1 n
Equiva ent y, 1 P ( En ) = im [1 P (En )] = 1 im P (En ) n=1 n n or P ( En ) = im P (En ) n=1 n Proposition 21.2 F is a nondecreasing function; that is, if a < b then F (a) F ( b). Proof. Suppose that a < b. Then {s : X(s) a} {s : X(s) b}. This implies that P (X a) P (X b). Hence, F (a) F (b)
192 DISCRETE RANDOM VARIABLES Example 21.1 Determine whether the given values can serve as the values of a dis tribution function of a random variable with the range x = 1, 2, 3, 4. F (1) = 0 .5, F (2) = 0.4, F (3) = 0.7, and F (4) = 1.0 Solution. No because F (2) < F (1) Proposition 21.3 F is continuous from the right. That is, limtb+ F (t) = F (b). Proof. Let {bn } be a decreasing sequence that converges to b with bn b for a n. De ne En = {s : X(s) bn }. Then {En }n1 is a decreasing sequence of events such that En = {s : X(s) b}. By Proposition 21.1 we have n=1 n im F (bn ) = im P (En ) = P ( En ) = P (X b) = F (b) n=1 n
Proposition 21.4 (a) imb F (b) = 0 (b) imb F (b) = 1 Proof. (a) Let {bn }n1 be a de reasing sequence with imn bn = . De ne En = {s : X(s) bn }. Then {En }n1 is a decre ng sequence of events such that En = {s : X(s) < }. By Proposition 21.1 we have n=1 n im F (bn ) = im P (En ) = P ( En ) = P (X < ) = 0. n=1 n
(b) Let {bn }n1 be an increasing sequence with imn bn = . De ne En = {s : X(s) bn }. Then {En }n1 is an increasing sequence of events such that En = {s : X(s) < }. By P roposition 21.1 we have n=1 n im F (bn ) = im P (En ) = P ( En ) = P (X < ) = 1 n=1 n Examp e 21.2 Determine whether the given va ues can serve as the va ues of a dis tribution function of a random variab e with the range x = 1, 2, 3, 4.
21 PROPERTIES OF THE CUMULATIVE DISTRIBUTION FUNCTION193 F (1) = 0.3, F (2) = 0. 5, F (3) = 0.8, and F (4) = 1.2 So ution. No because F (4) exceeds 1 A probabi ity questions can be answered in terms of the c.d.f. Proposition 21.5 P (X > a) = 1 F (a). Proof. We have P (X > a) = 1 P (X a) = 1 F (a) Examp e 21.3 Let X ha ve probabi ity mass function (pmf) p(x) = Find (a) the cumu ative distribution f unction (cdf) of X; (b) P (X > 5). So ution. (a) The cdf is given by F (x) = 0 1 [x] 8 1 8 for x = 1, 2, , 8. x<1 1x8 x>8 5 8 where [x] is the integer part of x. (b) We have P (X > 5) = 1 F (5) = 1 Proposit ion 21.6 P (X < a) = lim F n = 3 8 a 1 n = F (a ). Proof. We have P (X < a) =P n lim X a X a a 1 n 1 n = lim P n 1 n = lim F n
194 DISCRETE RANDOM VARIABLES Note that P (X < a) does not necessarily equal F (a), since F (a) also includes the probability that X equals a. Corollary 21.1 P (X a) = 1 im F n a 1 n = 1 F (a ). Proposition 21.7 If a < b then P (a < X b) = F (b) F (a). Proof. Let A = {s : X( s) > a} and B = {s : X(s) b}. Note that P (A B) = 1. Then P (a < X b) =P (A B) = P (A) + P (B) P (A B) =(1 F (a)) + F (b) 1 = F (b) F (a) Proposition 21.8 If a < b then P (a X < b) = imn F b 1 n
imn F a
1 n .
Proof. Let A = {s : X(s) a} and B = {s : X(s) < b}. Note that P (A B) = 1. Then using Proposition 21.6 we nd P (a X < b) =P (A B) =P (A) + P (B) P (A B) 1 1 + i m F b = 1 im F a n n n n 1 1 = im F b im F a n n n n Proposition 21. (a X b) = F (b) imn F a 1 1 n = F (b) F (a ).
21 PROPERTIES OF THE CUMULATIVE DISTRIBUTION FUNCTION195 Proof. Let A = {s : X(s ) a} and B = {s : X(s) b}. Note that P (A B) = 1. Then P (a X b) =P (A B) =P (A + P (B) P (A B) 1 = 1 im F a + F (b) 1 n n 1 =F (b) im F a n n Examp e that P (X = a) = F (a) F (a ). So ution. App ying the previous resu t we can writ e P (a x a) = F (a) F (a ) Proposition 21.10 If a < b then P (a < X < b) = imn F 1 n F (a). Proof. Let A = {s : X(s) > a} and B = {s : X(s) < b}. Note that P (A B) = 1. The n P (a < X < b) =P (A B) =P (A) + P (B) P (A B) =(1 F (a)) + im F n b 1 n 1 = im F n b 1 n F (a) Figure 21.2 i ustrates a typica F for a discrete random variab e X. Note that for a discrete random variab e the cumu ative distribution function wi a ways be a step function with jumps at each va ue of x that has probabi ity greater th an 0 and the size of the step at any of the va ues x1 , x2 , x3 , is equal to th e probability that X assumes that particular value.
196 DISCRETE RANDOM VARIABLES Figure 21.2
Example 21.5 (Mixed RV) The distribution function of a random variable X, is giv en by x<0 0, x , 0x<1 2 2 , 1x<2 F (x) = 3 11 , 2x<3 12 1, 3x (a) ompute P (X < 3). (c) Compute P (X = 1). (d) Compute P (X > 1 ) 2 (e) Compute P (2 < X 4).
Solution. (a) The graph is given in Figure 21.3. 1 1 (b) P (X < 3) = limn P X 3 n = limn F 3 n = 11 . 12 (c) P (X = 1) = P (X 1) P (X < 1) = F (1) limn F 1 2 1 3 6 (d) P (X > 1 ) = 1 P (X 1 ) = 1 F ( 1 ) = 3 . 2 2 2 4 1 (e) P (2 < X 4) = F (4) F (2) = 12 1 n =
21 PROPERTIES OF THE CUMULATIVE DISTRIBUTION FUNCTION197 Figure 21.3 Example 21.6 If X has the cdf forx < 1 0 1 4 for1 x < 1 1 for1 x < 3 F (x) = 2 3 for3 x < 5 4
nd (a) P (X 3) (b) P (X = 3) (c) P (X < 3) (d) P (X 1) (e) P (0.4 < X < 4) (f) P (0 .4 X < 4) (g) P (0.4 < X 4) (h) P (0.4 X 4) (i) P (X = 5). So ution. 3 (a) P (X = F (3) = 4 . 3 1 (b) P (X = 3) = F (3) F (3 ) = 4 2 = 1 (c) P (X < 3) = F (3 ) = 2 (d) P (X 1) = 1 F (1 ) = 1 1 = 3 4 4 1 4
(e) P (0.4 < X < 4) = F (4 ) F (0.4) = 3 1 = 1 4 4 2 1 3 (f) P (0.4 X < 4) = F (4 (0.4 ) = 4 1 = 2 4 3 1 1 (g) P (0.4 < X 4) = F (4) F (0.4) = 4 4 = 2 3 1 (h) P X 4) = F (4) F (0.4 ) = 4 4 = 1 2 (i) P (X = 5) = F (5) F (5 ) = 1 3 = 1 4 4 E e 21.7 The probabi ity mass function of X, the week y number of accidents at a c ertain intersection is given by p(0) = 0.40, p(1) = 0.30, p(2) = 0.20, and p(3) = 0.10. (a) Find the cdf of X. (b) Find the probabi ity that there wi be at e ast two accidents in any one week. So ution. (a) The cdf is given by forx < 0 0 0.40 for0 x < 1 0.70 for1 x < 2 F (x) = 0.90 for2 x < 3 1 forx 3 (b) P F (2 ) = 1 0.70 = 0.30
21 PROPERTIES OF THE CUMULATIVE DISTRIBUTION FUNCTION199
Practice Prob ems Prob em 21.1 In your pocket, you have 1 dime, 2 nicke s, and 2 pennies. You se e ct 2 coins at random (without rep acement). Let X represent the amount (in cents ) that you se ect from your pocket. (a) Give (exp icit y) the probabi ity mass f unction for X. (b) Give (exp icit y) the cdf, F (x), for X. (c) How much money d o you expect to draw from your pocket? Prob em 21.2 We are inspecting a ot of 2 5 batteries which contains 5 defective batteries. We random y choose 3 batteries . Let X = the number of defective batteries found in a samp e of 3. Give the cum u ative distribution function as a tab e. Prob em 21.3 Suppose that the cumu ati ve distribution 0 x 4 1 + x1 F (x) = 2 4 11 12 1 (a) Find P (X = i), (b) Find P ( 1 < X < 3 ). 2 2 Prob em 21.4 If the cumu ative distribution funct ion is given by x<0 0 1 0x<1 2 3 1x<2 5 F (x) = 4 5 2x<3 9 10 cu ate the probabi ity mass function. function is given by x<0 0x<1 1x<2 2x<3 3x
Prob em 21.5 Consider a random variab e X whose distribution function (cdf ) is given by x < 2 0 0.1 2 x < 1.1 0.3 1.1 x < 2 F (x) = 0.6 2x<3 abi ity mass function, p(x), of X, exp icit y. (b) Compute P (2 < X < 3). (c) Co mpute P (X 3). (d) Compute P (X 3 X 0). Prob em 21.6 Consider a random variab e X whose probabi ity mass function is given by p x = 1.9 0.1 x = 0.1 0.3 x x) = x=3 p 4p x=4 0 otherwise (a) What is p? (b) Find F (x) and sketch its gr . (c) What is F (0)? What is F (2)? What is F (F (3.1))? (d) What is P (2X 3 4 X 2.0)? (e) Compute E(F (X)). Prob em 21.7 The cdf of X is given by x < 4 0 0.3 < 1 F (x) = 0.7 1 x < 4 1 x4 (a) Find the probabi ity mass function. (b) Find th e variance and the standard deviation of X.
21 PROPERTIES OF THE CUMULATIVE DISTRIBUTION FUNCTION201 Prob em 21.8 In the gam e of dice ip, each p ayer ips a coin and ro s one die. If the coin comes up tai s, his score is the number of dots showing on the die. If the coin comes up heads, his score is twice the number of dots on the die. (i.e., (tai s,4) is worth 4 po ints, whi e (heads,3) is worth 6 points.) Let X be the rst p ayers score. (a) Find the probabi ity mass function p(x). (b) Compute the cdf F (x) for a numbers x . (c) Find the probabi ity that X < 4. Is this the same as F (4)? Prob em 21.9 A random variab e X has cumu ative distribution function x0 0 x2 0x<1 4 F (x) = 1 x < 2 4 1 x2 (a) What is the probabi ity that X = 0? What is the probabi ity t hat X = 1? What is the probabi ity that X = 2? 1 (b) What is the probabi ity tha t 2 < X 1? 1 (c) What is the probabi ity that 2 X < 1? (d) What is the probabi i ty that X > 1.5? Hint: You may nd it he pfu to sketch this function. Prob em 21. 10 Let X be a random variab e with the fo owing cumu ative distribution functio n: 0 x<0 1 2 x 0x< 2 F (x) = 1 x= 2 2x 12 x> 1 2 3 ( ) Find P (X > 2 ). (b) Find P ( 1 < X 3 ). 4 4 (c) Find . (d) Find P (X = 1 ) . 2 (e) Sketch the gr ph of F (x).
Continuous R ndom V ri bles Continuous r ndom v ri bles re r ndom qu ntities th t re me sured on continu ous sc le. They c n usu lly t ke on ny v lue over some interv l, which distingu ishes them from discrete r ndom v ri bles, which c n t ke on only sequence of v lues, usu lly integers. Typic lly r ndom v ri bles th t represent, for ex mple , time or dist nce will be continuous r ther th n discrete. 22 Distribution Functions We s y th t r ndom v ri ble is continuous if there exists nonneg tive functi on f (not necess rily continuous) de ned for ll re l numbers nd h ving the prope rty th t for ny set B of re l numbers we h ve P (X B) = B f (x)dx. We c ll the function f the prob bility density function ( bbrevi ted pdf) of the r ndom v ri ble X. If we let B = (, ) then f (x)dx = P [X (, )] = 1. Now, if we let B = [ , b] then b
f (x)dx. Th t is, re s under the prob bility density function represent prob bilities s illustr ted in Figure 22.1. 203
P (
X b) =
204 CONTINUOUS RANDOM VARIABLES
f (x)dx = 0.
F (t) = f (t)dt. Geometrica y, F (t) is the area under the graph of f to the eft of t. Examp e 22.1 If we think of an e ectron as a partic e, the function F (r) = 1 (2r2 + 2r + 1)e2r is the cumu ative distribution function of the distance, r, of the e ectr on in a hydrogen atom from the center of the atom. The distance is measured in B ohr radii. (1 Bohr radius = 5.29 1011 m.) Interpret the meaning of F (1).
It follows from this result th t P ( X < b) = P ( < X b) = P ( X b). nd P (X ) = P (X < ) nd P (X a) = P (X > a). The tion function (abbreviated cdf) F (t) of the random variab e X is de s F (t) = P (X t) i.e., F (t) is equa to the probabi ity that the sumes va ues, which are ess than or equa to t. From this de nition t
P (X =
) =
Figure 22.1 Now, if we let
= b in the previous formul we
nd
< X < b) = P ( cumu ative distribu ned as fo ow variab e X as we can write
22 DISTRIBUTION FUNCTIONS 205
So ution. F (1) = 1 5e2 0.32. This number says that the e ectron is within 1 Bohr radius from the center of the atom 32% of the time. Examp e 22.2 Find the distr ibution functions corresponding to the fo owing density functions: (a)f (x) = 1 , (1 + x2 ) ex (b)f (x) = , (1 + ex )2 a1 (c)f (x) = , (1 + x)a (d)f (x) =kx1 ek < < x < 0<x< 0 < x < , k > 0, > 0. Solution. ( ) x F (x) = 1 dy (1 + y 2 ) x 1 arctan y = 1 1 = arctan x 2 1 1 = arctan x + 2 (b) F (x) = ey dy y 2 1 + ey 1 = 1 + ex x
206 (c) For x 0 F (x) = CONTINUOUS RANDOM VARIABLES a1 dy a (1 + y) x 1 = (1 + y)a1 0 1 =1 (1 + x)a1 x For x < 0 it is obvious that F (x) = 0, so we could write the result in full as F (x) = (d) For x 0 x 0 1 1 (1+x)a1 x<0 x0 F (x) = 0 ky 1 eky dy = eky =1 ke For x < 0 we h ve F (x) = 0 so th t F (x) = x 0 kx 0 x<0 kx 1 ke x0
Next, we list the ro erties of the cumulative distribution function F (t) for a continuous random variable X. Theorem 22.1 The cumulative distribution function of a continuous random variable X satis es the following ro erties: (a) 0 F (t) 1. (b) F (t) = f (t). (c) F (t) is a non decreasing function, i.e. if a < b then F (a) F (b). (d) F (t) 0 as t and F (t) 1 as t . (e) P (a < X b) = F (b)
22 DISTRIBUTION FUNCTIONS 207 Proof. For (a), (c), (d), and (e) refer to Section 21. Part (b) is the result of a lying the Second Fundamental Theorem of Calculus to the function F (t) = t f (t)dt Figure 22.2 illustrates a re resentative cdf. Figure 22.2 Remark 22.1 The intuitive inter retation of the .d.f. is that for s mall a+ P (a X a + ) = F (a + ) F (a) = a f (t)dt f (a) In articular, P (a 2 X a + ) = f (a). 2
This means that the robability that X will be contained in an interval of lengt h around the oint a is a roximately f (a). Thus, f (a) is a measure of how lik ely it is that the random variable will be near a. Remark 22.2 By Theorem 22.1 ( b) and (d), limt f (t) = 0 = limt f (t). This follows from the fact that the gra h of F (t) levels o when t . That is, limt F (t) = 0. The density function for a contin s random variable , the model for some real life o ulation of data, will usuall y be a smooth curve as shown in Figure
208 22.3. CONTINUOUS RANDOM VARIABLES Figure 22.3 Exam le 22.3 Su ose that the function f (t) de ned below is the densi ty function of some random variable X. et t 0, f (t) = 0 t < 0. Com ute P (10 X 10 ). Solution. P (10 X 10) = = = = 0 10 10 10 f (t)dt 10 0 f (t)dt + f (t)dt 10 t e dt 0 et 0 = 1 e10 10 Example 22.4 Suppose the income (in tens of thousands of dollars) of people in a community can be approximated by a continuous distribution with density f (x) = 2x2 if x 2 0 if x < 2 Find the robability that a randomly chosen erson has an income between $30,000 and $50,000. Solution. Let X be the income of a randomly chosen erson. The ro bability that a
22 DISTRIBUTION FUNCTIONS randomly chosen erson has an income between $30,000 a nd $50,000 is 5 5 209 P (3 X 5) = 3 f (x)dx = 3 2x2 dx = 2x1 5 3 = 4 15
A df need not be continuous, as the following exam le illustrates. Exam le 22.5 (a) Determine the value of c so that the following function is a df. 15 x 64 + 64 2 x 0 3 + cx 0 < x 3 f (x) = 8 0 otherwise (b) Determine P (1 X 1). (c) F (x). Solution. (a) Observe that f is discontinuous at the oints 2 and 0, and is otentially also discontinuous at the oint 3. We rst nd the value of c that make s f a df. 0 1= 2 15 x + 64 64 2 0 3 dx + 0 3 + cx dx 8 3 0 3 15 x cx2 + x+ = x+ 64 128 2 8 2 30 2 9 9c = + + 64 64 8 2 100 9c + = 64 2 Solving for c we ws. 0 P (1 X 1) = 1 15 x + 64 64 1 dx + 0 3 x 8 8 nd c = 1 . 8 (b) The
robability P (1 X 1) is calculated as follo
dx = 69 128 (c) For 2 x 0 we have x F (x) = 2 15 t + 64 64 dt = x2 15 7 + x+ 128 64 16
210 and for 0 < x 3 0 CONTINUOUS RANDOM VARIABLES F (x) = 2 15 x + 64 64 x dx + 0 3 t 8 8 dt = 7 3 1 + x x2 . 16 8 16 Hence the full cdf is 0 x2 15 7 + 64 x + 16 128 7 3 1 + 8 x 16 x2 16 F (x) = 1 x < 2 2 x 0 0<x3 x>3 Observe that at all oints of discontinuity of the df, the cdf is continuous. T hat is, even when the df is discontinuous, the cdf is continuous
22 DISTRIBUTION FUNCTIONS 211 Practice Problems Problem 22.1 Determine the value of c so that the following function is a df. f (x) = c (x+1)3 0 if x 0 otherwise Problem 22.2 Let X denote the length of time (in minutes) that a certain young l ady s eaks on the tele hone. Assume that the df of X is given by f (x) = 1 x e 5 5 0 if x 0 otherwise (a) Find the robability that the young lady will talk on the tele hone for more than 10 minutes. (b) Find the robability that the young lady will talk on the tele hone between 5 and 10 minutes. (c) Find F (x). Problem 22.3 A college rofe ssor never nishes his lecture before the bell rings to end the eriod and always ishes his lecture within ten minutes after the bell rings. Su ose that the time X that ela ses between the bell and the end of the lecture has a robability de nsity function f (x) = kx2 0 x 10 0 otherwise
where k > 0 is a constant. Determine the value of k and nd the robability the le cture ends less than 3 minutes after the bell. Problem 22.4 Let X denote the lif etime (in hours) of a light bulb, and assume that the density function of X is g iven by 2x 0 x < 1 2 3 2<x<3 f (x) = 4 0 otherwise On average, what fraction of light bulbs last more than 15 minutes?
212 CONTINUOUS RANDOM VARIABLES Problem 22.5 De ne the function F : R R by 0 x<0 x/2 0x<1 F (x) = (x + 2)/6 1 1 x4 (a) Is F the cdf of a continuous random variable? Ex lain your answer. (b) If your answer to art (a) is yes, determine the corres onding df; if your answer was no, then make a modi cation to F so that it is a cdf, and then com ute the corr es onding df. Problem 22.6 The amount of time it takes a driver to react to a c hanging sto light varies from driver to driver, and is (roughly) described by th e continuous robability function of the form: f (x) = kxex x>0 0 otherwise where k is a constant and x is measured in seconds. (a) Find the constant k. (b) Find the robability that a driver takes more than 1 second to react. Problem 2 2.7 A random variable X has the cumulative distribution function ex . F (x) = x e +1 (a) Find the robability density function. (b) Find P (0 X 1). Problem 22.8 A lling station is su lied with gasoline once a week. If its weekly volume of s ales in thousands of gallons is a random variable with df f (x) = 5(1 x)4 0 < x < 1 0 otherwise What need the ca acity of the tank be so that the g exhausted in a given week is 0.1? robability of the su lys bein
22 DISTRIBUTION FUNCTIONS 213 Problem 22.9 The loss due to re in commerci l building is modeled by r ndom v ri ble X with density function f (x) = 0.005(20 x) 0 < x < 20 0 otherwise Given th t re loss exceeds 8, wh t is the prob bility th t it exceeds 16 ? Prob lem 22.10 The lifetime of m chine p rt h s continuous distribution on the in terv l (0, 40) with prob bility density function f, where f (x) is proportion l to (10 + x)2 . C lcul te the prob bility th t the lifetime of the m chine p rt is less th n 6. Problem 22.11 A group insur nce policy covers the medic l cl ims o f the employees of sm ll comp ny. The v lue, V, of the cl ims m de in one ye r is described by V = 100000Y where Y is r ndom v ri ble with density function f (x) = k(1 y)4 0 < y < 1 0 otherwise where k is const nt. Wh t is the condition l prob bility th t V exceeds 40,000 , given th t V exceeds 10,000? Problem 22.12 An insur nce comp ny insures l rg e number of homes. The insured v lue, X, of r ndomly selected home is ssumed to follow distribution with density function 3x4 x>1 f (x) = 0 otherwise Given th t r ndomly selected home is insured for t le st 1.5, wh t is the prob bili ty th t it is insured for less th n 2 ?
214 CONTINUOUS RANDOM VARIABLES Problem 22.13 An insur nce policy p ys for r ndom loss X subject to deductib le of C, where 0 < C < 1. The loss mount is modeled s continuous r ndom v ri ble with density function f (x) = 2x 0 < x < 1 0 otherwise
Let Y = m x{X1 , X2 , X3 }. Find P (Y > 1 ). 2 Problem 22.15 Let X h ve the dens ity function f (x) = 3x2 3 0<x< 0otherwise If P (X > 1) = 7 , nd the value of . 8 Problem 22.16 Insurance claims are for rand om amounts with a distribution function F (x). Su ose that claims which are no greater than a are no longer aid, but that the underlying distribution of claim s remains the same. Write down the distribution function of the claims aid, in terms of F (x). What is the distribution function for the claims not aid? Probl em 22.17 Su ose the robability function describing the life (in hours) of an o rdinary 60 Watt bulb is x 1 x>0 e 1000 1000 f (x) = 0 otherwise.
Given r ndom loss .5 is equ l to 0.64 endent, identic lly ) = 3x2 0 x 1 0
X, the prob bility th t the insur nce p yment is less th n 0 . C lcul te C. Problem 22.14 Let X1 , X2 , X3 be three indep distributed r ndom v ri bles e ch with density function f (x otherwise
22 DISTRIBUTION FUNCTIONS 215 Let A be the event that the bulb lasts 500 hours or more, and let B be the event that it lasts between 400 and 800 hours. (a) What is the robability of A? (b) What is the robability of B? (c) What is P (A B)?
216 CONTINUOUS RANDOM VARIABLES 23 Expectation, Variance and Standard Deviation As with discrete random variables, the expected value of a continuous random var iable is a measure of location. It de nes the balancing point of the distribution. Suppose that a continuous random variable X has a density function f (x) de ned i n [a, b]. Lets try to estimate E(X) by cutting [a, b] into n equal subintervals, each of width x, so x = (ba) . Let xi = a + ix, i = 0, 1, ..., n 1, n be the junctio n points between the subintervals. Then, the probability of X assuming a value o n [xi , xi+1 ] is xi+1 P (xi X xi+1 ) = xi f (x)dx xf xi + xi+1 2 where we used the midpoint rule to estimate the integral. An estimate of the des ired expectation is approximately n1 E(X) = i=0 xi + xi+1 2 xf xi + xi+1 2 . As n gets ever larger, we see that this converges to the integral b E(X) a xf (x)dx. The above argument applies if either a or b are in nite. In this case, one has to make sure that all improper integrals in question converge. From the above discu ssion it makes sense to de ne the expected value of X by the improper integral E(X) = xf (x)dx provided that the improper integral converges. Example 23.1 Find E(X) when the d ensity function of X is f (x) = 2x if 0 x 1 0 otherwise
23 EXPECTATION, VARIANCE AND STANDARD DEVIATION Solution. Using the formula for E(X) we nd 1 217 E(X) = xf (x)dx = 0 2x2 dx = 2 3 Example 23.2 The thickness of a conductive coating in micrometers has a density function of f (x) = 600x2 for 100m < x < 120m. Let X denote the coating thickness. (a) Determine the mean and variance of X. (b) If the coating costs $0.50 per mic rometer of thickness on each part, what is the average cost of the coating per p art? (c) Find the probability that the coating thickness exceeds 110m. Solution. (a) Note that f (x) = 600x2 for 100m < x < 120m and 0 elsewhere. So, by de nition, we have 120 E(X) = 100 x 600x2 dx = 600 ln x 120 109.39 100 and 120 = E(X ) (E(X)) = 100 2 2 2 x2 600x2 dx 109.392 33.19. (b) The average cost per part is $0.50 (109.39) = $54.70. (c) The desired probab ility is 120 P (X > 110) = 110 600x2 dx = 5 11 Sometimes for technical purposes the following theorem is useful. It expresses t he expectation in terms of an integral of probabilities. It is most often used f or random variables X that have only positive values; in that case the second te rm is of course zero. Theorem 23.1 Let Y be a continuous random variable with pr obability density function f. Then
0 E(Y ) = 0 P (Y > y)dy P (Y < y)dy.
218 CONTINUOUS RANDOM VARIABLES Proof. From the de nition of E(Y ) we have 0 E(Y ) = 0 xf (x)dx + y=x xf (x)dx 0 y=0 = 0 y=0 dyf (x)dx y=x dyf (x)dx Interchanging the order of integration as shown in Figure 23.1 we can write 0 y=x dyf (x)dx = y=0 0 y f (x)dxdy and 0 y=0 0 y dyf (x)dx = y=x f (x)dxdy. The result follows by putting the last two equations together and recalling that y f (x)dx = P (Y > y) and y f (x)dx = P (Y < y) Figure 23.1 Note that if X is a continuous random variable and g is a function d e ned for the values of X and with real values, then Y = g(X) is also a random
23 EXPECTATION, VARIANCE AND STANDARD DEVIATION 219 variable. The following theorem is particularly important and convenient. If a r andom variable Y = g(X) is expressed in terms of a continuous random variable, t hen this theorem gives the expectation of Y in terms of probabilities associated to X. Theorem 23.2 If X is continuous random variable with a probability densit y function f (x), and if Y = g(X) is a function of the random variable, then the expected value of the function g(X) is E(g(X)) = g(x)f (x)dx. Proof. By the previous theorem we have 0 E(g(X)) = 0 P [g(X) > y]dy P [g(X) < y]dy If we let By = {x : g(x) > y} then from the de nition of a continuous random varia ble we can write P [g(X) > y] = By f (x)dx = {x:g(x)>y} f (x)dx Thus, 0 E(g(X)) = 0 {x:g(x)>y} f (x)dx dy {x:g(x)<y} f (x)dx dy Now we can interchange the order of integration to obtain g(x) 0 E(g(X)) = {x:g(x)>0} 0 f (x)dydx {x:g(x)<0} g(x) f (x)dydx
= {x:g(x)>0} g(x)f (x)dx + {x:g(x)<0} g(x)f (x)dx = g(x)f (x)dx The following graph of g(x) might help understanding the process of interchangin g the order of integration that we used in the proof above
220 CONTINUOUS RANDOM VARIABLES Example 23.3 Assume the time T, measured in years, until failure of an appliance has density t 1 f (t) = e 10 , t 0. 10 The manufacturer o ers a warranty that ays an amount 100 if the a liance fails during the rst year, an amount 50 if it fai ls during the second or third year, and nothing if it fails after the third year . Determine the ex ected warranty ayment. Solution. Let X denote the warranty ayment by the manufacturer. We have 100 0 < T 1 50 1 < T 3 X= 0 T >3 Thus, 1 E(X) = 0 100 1 t e 10 dt + 10 1 10 1 3 50 1 1 10 3 1 t e 10 dt 10 3 =100(1 e ) + 50(e e 10 )
0 x>1 otherwise Wh t is the expected v lue of the bene t p id under the insur nce policy?
=100 50e 10 50e 10 Exam le 23.4 An insur nce policy reimburses loss up to limit of 10 . The policyholders loss, X, follows distribution with density fun ction: f (x) = 2 x3
bene
23 EXPECTATION, VARIANCE AND STANDARD DEVIATION Solution. Let Y denote the cl im p yments. Then Y = It follows th t E(Y ) = 2 2 10 3 dx dx + x3 x 1 10 10 10 2 2 = 1.9 = x 1 x 10 10 221 X 1 < X 10 10 X 10 x As a rst a lication of the revious theorem we have Corollary 23.1 For any const ants a and b E(aX + b) = aE(X) + b. Proof. Let g(x) = ax + b in the revious the orem to obtain E(aX + b) = =a (ax + b)f (x)dx xf (x)dx + b f (x)dx =aE(X) + b Exam le 23.5 Cl im mounts for wind d m ge to insured homes re indep endent r ndom v ri bles with common density function f (x) = 3 x4 0 x>1 otherwise where x is the mount of cl im in thous nds. Suppose 3 such cl ims will be m d e, wh t is the expected v lue of the l rgest of the three cl ims?
F (x) = 1 3 1 dt = 1 3 , x > 1 4 t x Next, let X1 , X2 , nd X3 denote the three cl ims m de th t h ve this distribut ion. Then if Y denotes the l rgest of these three cl ims, it follows th t the cd f of Y is given by FY (y) =P [(X1 y) (X2 y) (X3 y)] =P (X1 y)P (X2 y)P (X3 = 1 3 y 3 , y>1 The pdf of Y is obt ined by di erenti ting FY (y) 1 fY (y) = 3 1 3 y Fin lly, E(Y ) = 9 1 9 1 2 1 3 dy = 1 3 + 6 3 3 y y y y y 1 1 18 9 18 9 9 9 6 + 9 dy = 2 + 5 3 y y 2y 5y 8y 1 1 1 2 1 + =9 2.025 (in thous nds) 2 5 8 2 2 3 y4 = 9 y4 1 1 3 y 2 . dy
0 x > 0.6 otherwise To cover its losses, the m nuf cturer purch ses n insur nce policy with n nnu l deductible of 2. Wh t is the me n of the m nuf cturers nnu l losses not p id by the insur nce policy?
Ex mple 23.6 A m nuf cturers ction f (x) = 2.5(0.6)2.5 x3.5
nnu l losses follow
distribution with density fun
Solution. Note for x
ny of the r ndom v ri bles the cdf is given by
23 EXPECTATION, VARIANCE AND STANDARD DEVIATION Solution. Let Y denote the m nuf cturers ret ined nnu l losses. Then X 0.6 < X 2 Y = 2 X>2 Therefore, 2 2.5(0.6)2.5 2.5(0.6)2.5 2 dx + dx x3.5 x3.5 2 2.5(0.6)2.5 2(0.6)2.5 dx x2.5 x2 .5 2 2 223 E(Y ) = 0.6 2 x = 0.6 2(0.6)2.5 2.5(0.6)2.5 + 1.5x1.5 0.6 22.5 2.5(0.6)2.5 2.5(0.6)2.5 2(0.6)2.5 = + + 0.9343 1.5(2)1.5 1.5(0.6)1.5 22.5 = The v ri nce of r ndom v ri ble is me sur e of the spre d of the r ndom v ri ble bout its expected v lue. In essence, it te lls us how much v ri tion there is in the v lues of the r ndom v ri ble from its me n v lue. The v ri nce of the r ndom v ri ble X, is determined by c lcul ting the expect tion of the function g(X) = (X E(X))2 . Th t is, V r(X) = E (X E(X)) 2 . Theorem 23.3 ( ) An ltern tive formul for the v ri nce is given by V r(X) = E(X 2 ) [E(X)]2 . (b) For ny const nts nd b, V r( X + b) = 2 V r(X). Proo f. ( ) By Theorem 23.2 we h ve V r(X) = (x E(X))2 f (x)dx (x2 2xE(X) + (E(X))2 )f (x)dx = = x f (x)dx 2E(X) 2 xf (x)dx + (E(X)) 2 f (x)dx =E(X 2 ) (E(X))2
224 (b) We h ve CONTINUOUS RANDOM VARIABLES V r( X + b) = E[( X + b E( X + b))2 ] = E[ 2 (X E(X))2 ] = 2 V r(X) Ex mple 23. 7 Let X be r ndom v ri ble with prob bility density function f (x) = ( ) Find the v ri nce of X. (b) Find the c.d.f. F (x) of X. Solution. ( ) By the symmetry of the distribution bout 0, E(X) = 0. Thus, 0 1 2 4 x 1 < x < 2 , 2 0 elsewhere Var(X) =E(X ) = 1 2 1 2 2 x (2 + 4x)dx + 0 2 1 2 x2 (2 4x)dx =2 0 (2x2 8x3 )dx = 1 24 1 (b) Since the range of f is the interval ( 1 , 1 ), we have F (x) = 0 for x 2 2 2 1 1 and F (x) = 1 for x 2 . Thus it remains to consider the case when 2 < 1 x < 1 . For 2 < x 0, 2 x F (x) = 1 2 (2 + 4t)dt = 2x2 + 2x + 1 2 For 0 x < 1 , we have 2 0 x F (x) = 1 2 (2 + 4t)dt + 0 (2 4t)dt = 2x2 + 2x + 1 2 Combining these cases, we get 0 x < 1 + 1 0 x < 1 2 2 1 1 x 2 2 1 2x2 + 2x + 1 2 x < 0 2 F (x) =
2x
23 EXPECTATION, VARIANCE AND STANDARD DEVIATION Exam le 23.8 Let X be a continuo us random variable with df f (x) = 4xe2x x>0 0 otherwise n t t e dt 0 225 For this exam le, you might nd the identity (a) Find E(X). (b) Find the variance of X. (c) Find the robability that X < 1. Solution. (a) Using the substitution t = 2x we nd = n! useful. E(X) = 0 4x e 2 2x 1 dx = 2 t2 et dt = 0 2! =1 2 (b) First, we E(X 2 ) = 0 4x3 e2x dx = 1 4 t3 et dt = 0 3! 3 = 4 2 Hence, Var(X) = E(X 2 ) (E(X))2 = (c) We have 1 2 1 3 1= 2 2 P (X < 1) = P (X 1) = 0 4xe2x dx = 0 tet dt = (t + 1)et 2 0 nd E(X 2 ). Again, letting t = 2x we nd
= 13e2 As in the case of discrete random variable, it is easy to establish the formula Var(aX) = a2 Var(X) Exam le 23.9 Su ose that the cost of maintaining a car is g iven by a random variable, X, with mean 200 and variance 260. If a tax of 20% is introduced on all items associated with the maintenance of the car, what will t he variance of the cost of maintaining a car be?
226 CONTINUOUS RANDOM VARIABLES Solution. The new cost is 1.2X, so its variance is Var(1.2X) = 1.22 Var(X) = (1. 44)(260) = 374. Next, we de ne the standard deviation X to be the s uare root of t he variance. Exam le 23.10 A random variable has a Pareto distribution with ara meters > 0 nd x0 > 0 if its density function h s the form f (x) = x 0 x+1 0 x > x0 otherwise
f (x)dx = x0 x0 x0 x 0 dx = +1 x x =1 x0 (b) We h ve E(X) = x0 xf (x)dx = x0 x 0 dx = x 1 x 0 x1 = x0 x0 1 provided > 1. Simil rly, E(X ) = x0 2 x f (x)dx = x0 2
( ) Show th t f (x) is indeed density function. (b) Find E(X) tion. ( ) By de nition f (x) > 0. Also,
nd V r(X). Solu
x 0 dx = 1 x 2 x 0 x2 = x0 x2 0 2 provided > 2. Hence, V r(X) = x2 2 x2 x2 0 0 0 = 2 ( 1)2 ( 2)( 1)2
Percentiles nd Interqu rtile R nge We de ne the 100pth percentile of to be the qu ntity th t
popul tion
23 EXPECTATION, VARIANCE AND STANDARD DEVIATION 227 sep r tes the lower 100p% of popul tion from the upper 100(1 p)%. For r nd om v ri ble X, the 100pth percentile is the sm llest number x such th t F (x) . (23.3) The 50th ercentile is called the median of X. In the case of a continuous rando m variable, (23.3) holds with e uality; in the case of a discrete random variabl e, e uality may not hold. The di erence between the 75th ercentile and 25th erce ntile is called the inter uartile range. Exam le 23.11 Su ose the random variab le X has mf 1 (n) = 3 2 3 n , n = 0, 1, 2, Find the median and the 70th percentile. Solution. We have 1 F (0) = 0.33 3 1 2 F (1) = + 0.56 3 9 1 2 4 F (2) = + + 0.70 3 9 27 Thus, the median is 1 and the 7 0th percentile is 2 Example 23.12 Suppose the random variable X has pdf f (x) = Find the 100pth percentile. ex x0 0 otherwise
228 Solution. The cdf of X is given by CONTINUOUS RANDOM VARIABLES x F (x) = 0 et dt = 1 ex Thus, the 100 th ercentile is the solution to the e uation 1 ex = . For exam le , for the median, = 0.5 so that 1 ex = 0.5 Solving for x we nd x 0.693
23 EXPECTATION, VARIANCE AND STANDARD DEVIATION 229 Practice Problems Problem 23.1 Let X have the density function given by 0.2 1 < x 0 0.2 + cx 0 < x 1 f (x) = 0 otherwise (a) Find the value of c. (b) Find F (x). (c) Find P (0 x 0 .5). (d) Find E(X). Problem 23.2 The density function of X is given by f (x) = a + bx2 0 x 1 0 otherwise Su ose also that you are told that E(X) = 3 . 5 (a) Find a and b. (b) Determine the cdf, F (x), ex licitly. Problem 23.3 Com ute E(X) if X has the density func tion given by (a) x 1 x>0 xe 2 4 f (x) = 0 otherwise (b) f (x) = (c) f (x) = 5 x2 c(1 x2 ) 1 < x < 1 0 otherwise x>5 otherwise 0 Problem 23.4 A continuous random variable has a df f (x) = 1 0 x 2 0<x<2 otherwise Find the ex ected value and the variance.
230 CONTINUOUS RANDOM VARIABLES Problem 23.5 The lifetime (in years) of a certain brand of light bulb is a rando m variable (call it X), with robability density function f (x) = 4(1 + x)5 x0 0 o therwise (a) Find the mean and the standard deviation. (b) What is the robability that a randomly chosen lightbulb lasts less than a year? Problem 23.6 Let X be a conti nuous random variable with df f (x) = Find E(ln X). Problem 23.7 Let X have a c df F (x) = Find Var(X). Problem 23.8 Let X have a df f (x) = 1 1<x<2 0 otherwis e 1<x<e 0 otherwise 1 x 1 x1 1 x6 0 otherwise Find the ex ected value and variance of Y = X 2 . Problem 23.9 Let X be contin uous r ndom v ri ble with density function f (x) = C lcul te the expected v lue of X. x 10 0 2 x 4 otherwise
23 EXPECTATION, VARIANCE AND STANDARD DEVIATION 231 Problem 23.10 An auto insurance company insures an automobile worth 15,000 for o ne year under a policy with a 1,000 deductible. During the policy year there is a 0.04 chance of partial damage to the car and a 0.02 chance of a total loss of the car. If there is partial damage to the car, the amount X of damage (in thous ands) follows a distribution with density function f (x) = 0.5003e0.5x 0 < x < 15 0 otherwise What is the expected claim payment? Problem 23.11 An insurance companys monthly c laims are modeled by a continuous, positive random variable X, whose probability density function is proportional to (1 + x)4 , where 0 < x < . Determine the comp anys expected monthly claims. Problem 23.12 An insurers annual weather related los s, X, is a random variable with density function 2.5(200)2.5 x > 200 x3.5 f (x) = 0 otherwise Calculate the di erence between the 25th and 75th percentiles of X. Problem 23.13 A random variable X has the cumulative distribution function 0 x<1 x2 2x+2 F (x) = 1x<2 2 1 x2 Calculate the variance of X. Problem 23.14 A comp ny grees to ccept the highest of four se led bids on property. The four bids re reg rded s four independent r ndom v ri bles with common cumul tive distributi on function 1 3 5 F (x) = (1 + sin x), x 2 2 2 and 0 otherwise. What is the ex ecte d value of the acce ted bid?
232 CONTINUOUS RANDOM VARIABLES Problem 23.15 An insur nce policy on n electric l device p ys bene t of 4000 if the device f ils during the rst ye r. The mount of the bene t decre ses by 1000 e ch successive ye r until it re ches 0 . If the device h s not f iled by the beg inning of ny given ye r, the prob bility of f ilure during th t ye r is 0.4. Wh t is the expected bene t under this policy? Problem 23.16 Let Y be continuous r ndom v ri ble with cumul tive distribution function F (y) = 0 1e 1 (y )2 2 y otherwise
ex x>0 0 otherwise Prob em 23.18 Let f (x) be the density function of a random variab e X. We de ne t he mode of X to be the number that maximizes f (x). Let X be a continuous random variab e with density function f (x) = Find the mode of X. Prob em 23.19 A syst em made up of 7 components with independent, identica y distributed ifetimes w i operate unti any of 1 of the system s components fai s. If the ifetime X of each component has density function f (x) = 3 x4 x>1 0 otherwise 1 x(4 x) 0 < x < 3 9 0 otherwise What is the expected ifetime unti fai ure of the system?
where is const nt. Find the 75th percentile of Y. Problem 23.17 Let X be ndon v ri ble with density function f (x) = 1 Find if the median of X is 3 .
23 EXPECTATION, VARIANCE AND STANDARD DEVIATION Prob em 23.20 Let X have the den sity function f (x) = 2x 0 x k k2 0 otherwise 233 For what va ue of k is the variance of X equa to 2? Prob em 23.21 Peop e are di spersed on a inear beach with a density function f (y) = 4y 3 , 0 < y < 1, and 0 e sewhere. An ice cream vendor wishes to ocate her cart at the median of the ocations (where ha f of the peop e wi be on each side of her). Where wi she ocate her cart? Prob em 23.22 A game is p ayed where a p ayer generates 4 inde pendent Uniform(0,100) random variab es and wins the maximum of the 4 numbers. ( a) Give the density function of a p ayers winnings. (b) What is the p ayer expect ed winnings?
234 CONTINUOUS RANDOM VARIABLES 24 The Uniform Distribution Function The simp est continuous distribution is the uniform distribution. A continuous r andom variab e X is said to be uniform y distributed over the interva a x b if its pdf is given by f (x) = Since F (x) = x 1 ba 0 if a x b otherwise f (t)dt, the cdf is given by if x a 0 xa if a < x < b F (x) = ba 1 if x b
Figure 24.1 presents a graph of f (x) and F (x). Figure 24.1 If a = 0 and b = 1 then X is ca ed the standard uniform random vari ab e. Remark 24.1 The va ues at the two boundaries a and b are usua y unimporta nt because they do not a ter the va ue of the integra of f (x) over any interva . Some1 times they are chosen to be zero, and sometimes chosen to be ba . Our 1 de nition above assumes that f (a) = f (b) = f (x) = ba . In the case f (a) = f (b) = 0 then the pdf becomes f (x) = 1 ba 0 if a < x < b otherwise
24 THE UNIFORM DISTRIBUTION FUNCTION 235
f (x)dx = x d . ba Hence uniformity is the continuous equivalent of a discrete sample space in whic h every outcome is equally likely. Example 24.1 You are the production manager o f a soft drink bottling company. You believe that when a machine is set to dispe nse 12 oz., it really dispenses 11.5 to 12.5 oz. inclusive. Suppose the amount d ispensed has a uniform distribution. What is the probability that less than 11.8 oz. is dispensed? Solution. Since f (x) = 1 12.511.5 = 1, P (11.5 X 11.8) = area of rectangle of base 0.3 and height 1 = 0.3 Example 24.2 Suppose that X has a uniform distribution on the interval (0, a), where a > 0. F ind P (X > X 2 ). Solution. a 1 If a 1 then P (X > X 2 ) = 0 a dx = 1. If a > 1 then P (X > X 2 ) = 1 1 1 1 dx = a . Thus, P (X > X 2 ) = min{1, a } 0 a The exp ected value of X is b E(X) = a b xf (x) x dx ba b = a x2 = 2(b a) a b 2 a2 a+b = = 2(b a) 2
Because the pdf of a uniform random variab e is constant, if X is uniform, then the probabi ity X ies in any interva contained in (a, b) depends on y on the ength of the interva not ocation. That is, for any x and d such that [x, x + d ] [a, b] we have x+d
236 CONTINUOUS RANDOM VARIABLES and so the expected value of a uniform random variable is halfway between a and b. Because b E(X 2 ) = a x2 dx ba b x3 = 3(b a) a b 3 a3 = 3(b a) a2 + b2 + ab = 3 then Var(X) = E(X 2 ) (E(X))2 = a 2 + b2 + ab (a + b)2 (b a)2 = . 3 4 12 Example 24.3 You arrive at a bus stop 10:00 am, knowing that the bus will arrive at some time uniformly distributed between 10:00 and 10:30 am. Let X be your wa it time for the bus. Then X is uniformly distributed in (0, 30). Find E(X) and V ar(X). Solution. We have a = 0 and b = 30. Thus, E(X) = 302 = 75 12 a+b 2 (ba)2 12 = 15 and Var(X) = =
24 THE UNIFORM DISTRIBUTION FUNCTION 237 Practice Problems Problem 24.1 The total time to process a loan application is uniformly distribut ed between 3 and 7 days. (a) Let X denote the time to process a loan application . Give the mathematical expression for the probability density function. (b) Wha t is the probability that a loan application will be processed in fewer than 3 d ays ? (c) What is the probability that a loan application will be processed in 5 days or less ? Problem 24.2 Customers at TAB are charged for the amount of sala d they take. Sampling suggests that the amount of salad taken is uniformly distr ibuted between 5 ounces and 15 ounces. Let X = salad plate lling weight (a) Find the probability density function of X. (b) What is the probability that a custom er will take between 12 and 15 ounces of salad? (c) Find E(X) and Var(X). Proble m 24.3 Suppose thst X has a uniform distribution over the interval (0, 1). Find (a) F (x). (b) Show that P (a X a + b) for a, b 0, a + b 1 de ends only on b. Pr oblem 24.4 Let X be uniform on (0,1). Com ute E(X n ). Problem 24.5 Let X be a u niform random variable on the interval (1,2) and let Y = Find E[Y ]. 1 . X Problem 24.6 You arrive at a bus sto at 10:00 am, knowing that the bus will arr ive at some time uniformly distributed between 10:00 and 10:30. What is the rob ability that you will have to wait longer than 10 minutes?
238 CONTINUOUS RANDOM VARIABLES Problem 24.7 An insur nce policy is written to cover loss, X, where X h s un iform distribution on [0, 1000]. At wh t level must deductible be set in order for the expected p yment to be 25% of wh t it would be with no deductible? Prob lem 24.8 The w rr nty on m chine speci es th t it will be repl ced t f ilure or ge 4, whichever occurs rst. The m chines ge t f ilure, X, h s density function 1 0<x<5 5 f (x) = 0 otherwise Let Y be the ge of the m chine t the time of re pl cement. Determine the v ri nce of Y. Problem 24.9 The owner of n utomobile insures it g inst d m ge by purch sing n insur nce policy with deductible of 250 . In the event th t the utomobile is d m ged, rep ir costs c n be modeled by uniform r ndom v ri ble on the interv l (0, 1500). Determine the st nd rd d evi tion of the insur nce p yment in the event th t the utomobile is d m ged. P roblem 24.10 Let X be r ndom v ri ble distributed uniformly over the interv l [1, 1]. ( ) Compute E(eX ). (b) Compute V r(eX ). Problem 24.11 Let X be r ndom v ri ble with continuous uniform distribution on the interv l (1, ), > 1. If E(X) = 6V r(X), wh t is the v lue of ? Problem 24.12 Let X be r ndom v ri bl e with continuous uniform distribution on the interv l (0, 10). Wh t is P (X + 10 > 7)? X Problem 24.13 Suppose th t X h s uniform distribution on the inter v l (0, ), > 0. Find P (X > X 2 ).
25 NORMAL RANDOM VARIABLES 239
< x < .
Figure 25.1 The normal di tribution i u ed to model phenomenon uch a a per on height at a certain age or the mea urement error in an experiment. Ob erve that the di tribution i ymmetric about the point hence the experiment outcome being modeled hould be equaly likely to a ume point above a point below . The norm al di tribution i probably the mo t important di tribution becau e of a re ult we will ee later, known a the central limit theorem. To prove that the given f (x) i indeed a pdf we mu t how that the area under the normal curve i 1. Tha t i , (x)2 1 e 22 dx = 1. 2 Fir t note that u ing the ub titution y = x we have (x)2 1 1 e 22 dx = 2 2 e 2 dy. y2
Thi den ity function i a bell haped curve that i 25.1).
25 Norm l R ndom V ri bles A norm l r ndom v ri ble with p r meters nd 2 ha (x)2 1 e 22 , 2
a pdf f (x) =
ymmetric about (See Figure
240 Toward this end, let I = y e 2 2 CONTINUOUS RANDOM VARIABLES dy. Then y2 I2 = e 2 dy 2 0 0 e 2 dx dxdy x2 = e 2 2 x +y 2 = 0 e 2 rdrd re 2 dr = 2 r2 r2 =2 Thus, I = 2 and the result is roved. Note that in the rocess above, we used the olar substitution x = r cos , y = r sin , and dydx = rdrd. Exam le 25.1 The width of a bolt of fabric is normally distributed with mean 950mm and standard deviat ion 10 mm. What is the robability that a randomly chosen bolt has a width betwe en 947 and 950 mm? Solution. Let X be the width (in mm) of a randomly chosen bol t. Then X is normally distributed with mean 950 mm and variation 100 mm. Thus, 1 P (947 X 950) = 10 2 950 e 947 (x950)2 200 dx 0.118 Theorem 25.1 If X is a normal distribution with arameters (, 2 ) then Y = aX + b i a normal di tribution with parmater (a + b, a2 2 ). Proof. We prove the re u lt when a > 0. The proof i imilar for a < 0. Let FY denote the cdf of Y. Then FY (x) =P (Y x) = P (aX + b x) xb xb =P X = FX a a
25 NORMAL RANDOM VARIABLES Di erentiating both ide to obtain xb 1 fY (x) = fX a a 1 xb = exp ( )2 /(2 2 ) a 2a 1 = exp (x (a + b))2 /2(a)2 2a 241 which how that Y i normal with parameter (a + b, a2 2 ) Note that if Z = X then thi i a normal di tribution with parameter (0,1). Such a random variable i called the tandard normal random variable. Theorem 25.2 If X i a normal random variable with parameter (, 2 ) then (a) E(X) = (b) Var(X) = 2 . Proof. (a) Let Z = E(Z) = Thu , E(X) = E(Z + ) = E(Z) + = . (b) 1 Var(Z) = E(Z 2 ) = 2 x2 X be the tandard normal di tribution. Then 1 2 xfZ (x)dx = xe 2 dx = 1 e 2 2 x2 x2 =0 x2 e 2 dx. x2 Using integration by arts with u = x and dv = xe 2 we nd Var(Z) = Thus, Var(X) = Var(Z + ) = 2 Var(Z) = 2 1 2 xe 2 x2 + x2 e 2 dx = 1 2 x2 e 2 dx = 1.
242 CONTINUOUS RANDOM VARIABLES Figure 25.2 shows di erent normal curves with the same and di erent . Figure 25.2 Example 25.2 A college ha an enrollment of 3264 female tudent . Re cord how that the mean height of the e tudent i 64.4 inche and the tandar d deviation i 2.4 inche . Since the hape of the relative hi togram of ample c ollege tudent approximately normally di tributed, we a ume the total populati on di tribution of the height X of all the female college tudent follow the n ormal di tribution with the ame mean and the tandard deviation. Find P (66 X 6 8). Solution. If you want to nd out the percentage of tudent who e height are between 66 and 68 inche , you have to evaluate the area under the normal curve f rom 66 to 68 a hown in Figure 25.3. Figure 25.3
25 NORMAL RANDOM VARIABLES Thu , 1 P (66 X 68) = 2(2.4) 68 243 e 66 (x64.4)2 2(2.4)2 dx 0.1846 It is traditional to denote the cdf of Z by (x). That is, 1 (x) = 2 2 x 2 x e 2 dy. y2
Now, since fZ (x) = (x) = 1 e then fZ (x) is an even function. This 2 im lies that (x) = (x). Integrating we nd that (x) = (x) + C. Letting x = 0 we nd that C = 2 0.5) = 1. Thus, (x) = 1 (x), This im lies that P (Z x) = P (Z > x). Now, (x) is th rea under the standard curve to the left of x. The values of (x) for x 0 are give n in Table 25.1 below. Equation 25.4 is used for x < 0. Exam le 25.3 On May 5, i n a certain city, tem eratures have been found to be normally distributed with m ean = 24 C and variance 2 = 9. The record temperature on that day i 27 C. (a) Wha t is the robability that the record of 27 C will be broken next May 5? (b) What is the robability that the record of 27 C will be broken at least 3 times during the next 5 years on May 5 ? (Assume that the tem eratures during the next 5 yea rs on May 5 are inde endent.) (c) How high must the tem erature be to lace it a mong the to 5% of all tem eratures recorded on May 5? Solution. (a) Let X be th e tem erature on May 5. Then X has a normal distribution with = 24 and = 3. The de ired probability i given by P (X > 27) =P 27 24 X 24 > = P (Z > 1) 3 3 =1 P (Z 1) = 1 (1) = 1 0.8413 = 0.1587 < x < . (25.4)
244 CONTINUOUS RANDOM VARIABLES (b) Let Y be the number of times with broken records during the next 5 years on May 5. Then, Y has a binomial distribution with n = 5 and = 0.1587. So, the de sired robability is P (Y 3) =P (Y = 3) + P (Y = 4) + P (Y = 5) =C(5, 3)(0.1587) 3 (0.8413)2 + C(5, 4)(0.1587)4 (0.8413)1 +C(5, 5)(0.1587)5 (0.8413)0 0.03106 (c) Let x be the desired tem erature. We must have P (X > x) = 0.05 or equivalently P (X x) = 0.95. Note that P (X x) = P X 24 x 24 < 3 3 =P Z< x 24 3 = 0.95 rom the normal Table (at the end of this section) we nd P (Z 1.65) = 0.95. Thus, we set x24 = 1.65 and solve for x we nd x = 28.95 C 3 Next, we oint out that rob abilities involving normal random variables are reduced to the ones involving st andard normal variable. or exam le P (X a) = P a X = a .
Example 25.4 Let X be a normal random variable with parameter and 2 . Find (a)P ( X + ). (b)P ( 2 X + 2). (c)P ( 3 X + 3). Solution. (a) ) =2(0.8413) 1 = 0.6826. Thus, 68.26% of all ossible observations lie within on e standard deviation to either side of the mean.
25 NORMAL RANDOM VARIABLES (b) We have P ( 2 X + 2) =P (2 Z 2) =(2) (2 0.9544. 245 Thus, 95.44% of all ossible observations lie within two standard deviations to either side of the mean. (c) We have P ( 3 X + 3) =P (3 Z 3) =(3) (3) 9974. Thus, 99.74% of all ossible observations lie within three standard deviat ions to either side of the mean The Normal A roximation to the Binomial Distribution Historically, the normal d istribution was discovered by DeMoivre as an a roximation to the binomial distr ibution. The result is the so called De Moivre La lace theorem. Theorem 25.3 Let Sn denote the number of successes that occur with n inde endent Bernoulli
246 CONTINUOUS RANDOM VARIABLES trials, each with robability ) b = (b) (a). of success. Then, for a < b, lim P a Sn n n (1
n Proof. This result is a s ecial case of the central limit theorem, which will be discussed in Section 40. Consequently, we will defer the roof of this result u ntil then Remark 25.1 The normal a roximation to the binomial distribution is u sually quite good when n 10 and n(1 ) 10. Remark 25.2 Su ose we are a roxima ting a binomial random variable with a normal random variable. Say we want to nd P (X 10). igure 25.4 zooms in on the left ortion of the distributions and show s P (X 10). Since each histogram bar is centered at an integer, the shaded area actually goes to 10.5. If we were to a roximate P (X 10) with the normal cdf ev aluated at 10, we would e ectively be missing half of the bar centered at 10. In ractice, then, we a ly a continuity correction, by evaluating the normal cdf at 10.5. igure 25.4
Exam le 25.5 A rocess yields 10% defective items. If 100 items are randomly sel ected from the rocess, what is the robability that the number of defectives ex ceeds 13? Solution. Let X be the number of defective items. Then X is binomial w ith n = 100 and = 0.1. Since n = 10 10 and n(1 ) = 90 10 we can use the norm al a roximation to the binomial with = n = 10 and 2 = np(1 p) = 9. We want P ( X > 13). U ing continuity correction we nd P (X > 13) =P (X 14) 13.5 10 X 10 ) = ( 9 9 1 (1.17) = 1 0.8790 = 0.121 Exam le 25.6 In the United States, 1 of the eo le are lefthanded. In a small town (a 6 random sam le) of 612 ersons, estimate the robability that the number of lefthanded ersons is strictly between 90 an d 150. Solution. Let X be the number of left handed eo le in the sam le. Then X is a binomial random variable with n = 612 and = 1 . Since n = 102 10 and 6 n(1 ) = 510 10 we can use the normal a roximation to the binomial with = n = 102 and 2 = np(1 p) = 85. U ing continuity correction we nd P (90 < X < 150) =P ( 91 X 149) = 90.5 102 X 102 149.5 102 =P 85 85 85 =P (1.25 Z 5.15) number i found by u ing TI83 Plu Example 25.7 There are 90 tudent in a tat i tic cla . Suppo e each tudent ha a tandard deck of 52 card of hi /her ow n, and each of them elect 13 card at
248 CONTINUOUS RANDOM VARIABLES random without replacement from hi /her own deck independent of the other . What i the chance that there are more than 50 tudent who got at lea t 2 or more a ce ? Solution. Let X be the number of tudent who got at lea t 2 ace or more, then clearly X i a binomial random variable with n = 90 and p= C(4, 2)C(48, 11 ) C(4, 3)C(48, 10) C(4, 4)C(48, 9) + + 0.2573 C(52, 13) C(52, 13) C(52, 13)
Since np 23.573 10 and n(1 p) = 66.843 10, X can be approximated by a normal ran dom variable with = 23.573 and = np(1 p) 4.1473. Thu , P (X > 50) =1 P (X 50) = 1 1 (6.5) 50.5 23.573 4.1473
25 NORMAL RANDOM VARIABLES Area under the Standard Normal Curve from to x x 0.0 0 .1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 2 .1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0 3.1 3.2 3.3 3.4 0.00 0.5000 0.5398 0.5793 0.6179 0.6554 0.6915 0.7257 0.7580 0.7881 0.8159 0.8413 0.8643 0.8849 0.9032 0. 9192 0.9332 0.9452 0.9554 0.9641 0.9713 0.9772 0.9821 0.9861 0.9893 0.9918 0.993 8 0.9953 0.9965 0.9974 0.9981 0.9987 0.9990 0.9993 0.9995 0.9997 0.01 0.5040 0.5 438 0.5832 0.6217 0.6591 0.6950 0.7291 0.7611 0.7910 0.8186 0.8438 0.8665 0.8869 0.9049 0.9207 0.9345 0.9463 0.9564 0.9649 0.9719 0.9778 0.9826 0.9864 0.9896 0. 9920 0.9940 0.9955 0.9966 0.9975 0.9982 0.9987 0.9991 0.9993 0.9995 0.9997 0.02 0.5080 0.5478 0.5871 0.6255 0.6628 0.6985 0.7324 0.7642 0.7939 0.8212 0.8461 0.8 686 0.8888 0.9066 0.9222 0.9357 0.9474 0.9573 0.9656 0.9726 0.9783 0.9830 0.9868 0.9898 0.9922 0.9941 0.9956 0.9967 0.9976 0.9982 0.9987 0.9991 0.9994 0.9995 0. 9997 0.03 0.5120 0.5517 0.5910 0.6293 0.6664 0.7019 0.7357 0.7673 0.7967 0.8238 0.8485 0.8708 0.8907 0.9082 0.9236 0.9370 0.9484 0.9582 0.9664 0.9732 0.9788 0.9 834 0.9871 0.9901 0.9925 0.9943 0.9957 0.9968 0.9977 0.9983 0.9988 0.9991 0.9994 0.9996 0.9997 0.04 0.5160 0.5557 0.5948 0.6331 0.6700 0.7054 0.7389 0.7704 0.79 95 0.8264 0.8508 0.8729 0.8925 0.9099 0.9251 0.9382 0.9495 0.9591 0.9671 0.9738 0.9793 0.9838 0.9875 0.9904 0.9927 0.9945 0.9959 0.9969 0.9977 0.9984 0.9988 0.9 992 0.9994 0.9996 0.9997 0.05 0.5199 0.5596 0.5987 0.6368 0.6736 0.7088 0.7422 0 .7734 0.8023 0.8289 0.8531 0.8749 0.8944 0.9115 0.9265 0.9394 0.9505 0.9599 0.96 78 0.9744 0.9798 0.9842 0.9878 0.9906 0.9929 0.9946 0.9960 0.9970 0.9978 0.9984 0.9989 0.9992 0.9994 0.9996 0.9997 0.06 0.5239 0.5636 0.6026 0.6406 0.6772 0.712 3 0.7454 0.7764 0.8051 0.8315 0.8554 0.8770 0.8962 0.9131 0.9279 0.9406 0.9515 0 .9608 0.9686 0.9750 0.9803 0.9846 0.9881 0.9909 0.9931 0.9948 0.9961 0.9971 0.99 79 0.9985 0.9989 0.9992 0.9994 0.9996 0.9997 249 0.07 0.5279 0.5675 0.6064 0.6443 0.6808 0.7157 0.7486 0.7794 0.8078 0.8340 0.857 7 0.8790 0.8980 0.9147 0.9292 0.9418 0.9525 0.9616 0.9693 0.9756 0.9808 0.9850 0 .9884 0.9911 0.9932 0.9949 0.9962 0.9972 0.9979 0.9985 0.9989 0.9992 0.9995 0.99 96 0.9997 0.08 0.5319 0.5714 0.6103 0.6480 0.6844 0.7190 0.7517 0.7823 0.8106 0.8365 0.859 9 0.8810 0.8997 0.9162 0.9306 0.9429 0.9535 0.9625 0.9699 0.9761 0.9812 0.9854 0 .9887 0.9913 0.9934 0.9951 0.9963 0.9973 0.9980 0.9986 0.9990 0.9993 0.9995 0.99 96 0.9997 0.09 0.5359 0.5753 0.6141 0.6517 0.6879 0.7224 0.7549 0.7852 0.8133 0.8389 0.862 1 0.8830 0.9015 0.9177 0.9319 0.9441 0.9545 0.9633 0.9706 0.9767 0.9817 0.9857 0 .9890 0.9916 0.9936 0.9952 0.9964 0.9974 0.9981 0.9986 0.9990 0.9993 0.9995 0.99 97 0.9998
250 CONTINUOUS RANDOM VARIABLES Practice Problems Problem 25.1 Scores for a articular standardized test are normally distributed with a mean of 80 and a standard deviation of 14. ind the robability that a ra ndomly chosen score is (a) no greater than 70 (b) at least 95 (c) between 70 and 95. (d) A student was told that her ercentile score on this exam is 72%. A ro ximately what is her raw score? Problem 25.2 Su ose that egg shell thickness is normally distributed with a mean of 0.381 mm and a standard deviation of 0.031 mm. (a) ind the ro ortion of eggs with shell thickness more than 0.36 mm. (b) ind the ro ortion of eggs with shell thickness within 0.05 mm of the mean. (c) ind the ro ortion of eggs with shell thickness more than 0.07 mm from the mea n. Problem 25.3 Assume the time required for a certain distance runner to run a mile follows a normal distribution with mean 4 minutes and variance 4 seconds. ( a) What is the robability that this athlete will run the mile in less than 4 mi nutes? (b) What is the robability that this athlete will run the mile in betwee n 3 min55sec and 4min5sec? Problem 25.4 You work in Quality Control for GE. Ligh t bulb life has a normal distribution with = 2000 hours and = 200 hour . What e probability that a bulb will la t (a) between 2000 and 2400 hour ? (b) le th an 1470 hour ? Problem 25.5 Human intelligence (IQ) ha been hown to be normall y di tributed with mean 100 and tandard deviation 15. What fraction of people h ave IQ greater than 130 (the gifted cuto ), given that (2) = .9772?
th
25 NORMAL RANDOM VARIABLES 251 Problem 25.6 Let X re resent the lifetime of a randomly chosen battery. Su ose X is a normal random variable with arameters (50, 25). (a) ind the robability that the battery lasts at least 42 hours. (b) ind the robability that the bat tery will lasts between 45 to 60 hours. Problem 25.7 Su ose that 25% of all lic ensed drivers do not have insurance. Let X be the number of uninsured drivers in a random sam le of 50. (a) What is the robability that the number of uninsured drivers is at most 10? (b) What is the robability that the number of uninsured drivers is between 5 and 15? Problem 25.8 For Comp ny A there is 60% ch nce t h t no cl im is m de during the coming ye r. If one or more cl ims re m de, the tot l cl im mount is norm lly distributed with me n 10,000 nd st nd rd devi t ion 2,000 . For Comp ny B there is 70% ch nce th t no cl im is m de during the coming ye r. If one or more cl ims re m de, the tot l cl im mount is norm lly distributed with me n 9,000 nd st nd rd devi tion 2,000 . Assuming th t the to t l cl im mounts of the two comp nies re independent, wh t is the prob bility th t, in the coming ye r, Comp ny Bs tot l cl im mount will exceed Comp ny As tot l cl im mount? Problem 25.9 If for cert in norm l r ndom v ri ble X, P (X < 500) = 0.5 nd P (X > 650) = 0.0227, nd the st nd rd devi tion of X. Problem 25.1 0 A computer gener tes 10, 000 r ndom decim l digits 0, 1, 2, , 9 (each se1 lect ed with probability 10 ) and then tabulates the number of occurrences of each di git. (a) Using an appropriate approximation, nd the (approximate) probability tha t exactly 1,060 of the 10,000 digits are 0. (b) Determine x (as small as possibl e) such that, with 99.95% probability, the number of digits 0 is at most x.
252 CONTINUOUS RANDOM VARIABLES Problem 25.11 Suppose that X is a normal random variable with parameters = 5, 2 = 49. U ing the table of the normal di tribution , compute: (a) P (X > 5.5), (b) P (4 < X < 6.5), (c) P (X < 8), (d) P ( X 7 4). Problem 25.12 A com any wants to buy boards of length 2 meters and is willing to acce t lengths that are o by a s much as 0.04 meters. The board manufacturer roduces boards of length normally distributed with mean 2.01 meters and standard deviation . If the probability th at a board i too long i 0.01, what i ? Problem 25.13 Let X be a normal random variable with mean 1 and variance 4. Find P (X 2 2X 8). Problem 25.14 Score on a tandardized exam are normally di tributed with mean 1000 and tandard deviati on 160. (a) What proportion of tudent core under 850 on the exam? (b) They wi h to calibrate the exam o that 1400 repre ent the 98th percentile. What houl d they et the mean to? (without changing the tandard deviation) Problem 25.15 The daily number of arrival to a rural emergency room i a Poi on random varia ble with a mean of 100 people per day. U e the normal approximation to the Poi on di tribution to obtain the approximate probability that 112 or more people ar rive in a day. Problem 25.16 A machine i u ed to automatically ll 355ml pop bott le . The actual amount put into each bottle i a normal random variable with mea n 360ml and tandard deviation of 4ml. (a) What proportion of bottle are lled wi th le than 355ml of pop? b) Suppo e that the mean ll can be adju ted. To what v alue hould it be et o that only 2.5% of bottle are lled with le than 355ml?
Problem 25.17 Suppo e that your journey time from home to campu i normally di tributed with mean equal to 30 minute and tandard deviation equal to 5 minute . What i the late t time that you hould leave home if you want to be over 99% ure of arriving in time for a cla at 2pm? Problem 25.18 Suppo e that a local vote i being held to ee if a new manufacturing facility will be built in the l ocality. A polling company will urvey 200 individual to mea ure upport for th e new facility. If in fact 53% of the population oppo e the building of thi fac ility, u e the normal approximation to the binomial, with a continuity correctio n, to approximate the probability that the poll will how a majority in favor? P roblem 25.19 Suppo e that the current mea urement in a trip of wire are a ume d to follow a normal di tribution with mean of 12 milliampere and a tandard de viation of 3 (milliampere ). (a) What i the probability that a mea urement will exceed 14 milliampere ? (b) What i the probability that a current mea urement i between 9 and 16 milliampere ? (c) Determine the value for which the probabil ity that a current mea urement i below thi value i 0.95.
254 CONTINUOUS RANDOM VARIABLES 26 Exponential Random Variable An exponential random variable with parameter > 0 is a random variab e with pdf ex if x 0 f (x) = 0 if x < 0 Note that 0 ex dx = ex 0 =1 The graph of the probabi ity density function is shown in Figure 26.1 Figure 26.1 Exponentia random variab es are often used to mode arriva times, waiting times, and equipment fai ure times. The expected va ue of X can be found using integration by parts with u = x and dv = ex dx : E(X) = 0 xex dx 0 = xex = + 0 ex dx 0 xex 0 1 + ex 1 =
26 EXPONENTIAL RANDOM VARIABLES Furthermore, using integration by parts again, w e may a so obtain that 255 E(X 2 ) = 0 x2 ex dx = x2 ex 0 0 x2 d(ex ) xex dx o = +2 2 = 2 Thus, Var(X) = E(X 2 ) (E(X))2 = 1 1 2 2 = 2. 2 Examp e 26.1 The time between machine fai ures at an industria p ant has an exp onentia distribution with an average of 2 days between fai ures. Suppose a fai ure has just occurred at the p ant. Find the probabi ity that the next fai ure w ont happen in the next 5 days. So ution. Let X denote the time between accidents. The mean time to fai ure is 2 days. Thus, = 0.5. Now, P (X > 5) = 1 P (X 5) = 5 0.5e0.5x dx 0.082085 Examp e 26.2 The mi eage (in thousands of mi es) which car owners get with a certain kind of radia tire is a random variab e having an exp onentia distribution with mean 40. Find the probabi ity that one of these tires wi ast at most 30 thousand mi es. So ution. Let X denote the mi eage (in tho usdands of mi es) of one these tires. Then 1 X is an exponentia distribution wi th paramter = 40 . Thus, 30 P (X 30) = 0 1 x e 40 dx 40 x = e 40 x 30 0 = 1 e 4 0.5276 3 The cumu ative distribution function is given by F (x) = P (X x) = 0 eu du = eu x = 1 ex , x 0 0
Examp e 26.3 Suppose that the ength of a phone ca in minutes is an exponentia random variab e with mean 10 minutes. If someone arrives immediate y ahead of you at a pub ic te ephone booth, nd the probabi ity that you have to wait (a) mor e than 10 minutes (b) between 10 and 20 minutes So ution. Let X be the time you must wait at the phone booth. Then X is an exponentia random variab e with para meter = 0.1. (a) We have P (X > 10) = 1 F (10) = 1 (1 e1 ) = e1 0.3679. (b) We e P (10 X 20) = F (20) F (10) = e1 e2 0.2325 The most important property of the ponentia distribution is known as the memory ess property: P (X > s + t X > s) = P (X > t), s, t 0. This says that the robability that we have to wait for an additional time t (an d therefore a total time of s + t) given that we have already waited for time s is the same as the robability at the start that we would have had to wait for t ime t. So the ex onential distribution forgets that it is larger than s. To see wh y the memoryless ro erty holds, note that for all t 0, we have P (X > t) = t ex dx = ex = et . t It fo ows that P (X > s + t X > s) = P (X > s + t and X > s) P (X > s) P (X > s + t) = P (X > s) (s+t) e = s e =et = P (X > t) Examp e 26.4 Let X be the time (in hours) required to repair a computer system. We
26 EXPONENTIAL RANDOM VARIABLES 257 assume that X has an exponentia distribution with parameter = 1 . Find 4 (a) th e cumu ative distribution function of X. (b) P (X > 4). (c) P (X > 10 X > 8). So lution. (a) It is easy to see that the cumulative distribution function is F (x) = 1 e 4 x0 0 elsewhere 4 x
(b) P (X > 4) = 1 P (X 4) = 1 (4) = 1 (1 e 4 ) = e1 0.368. (c) By the memor ro erty, we nd P (X > 10 X > 8) =P (X > 8 + 2 X > 8) = P (X > 2) =1 P (X 2) = 1 F (2) =1 (1 e 2 ) = e 2 0.6065 Example 26.5 The distance between major cracks in a highway follows an exponential distribution with a mean of 5 miles. (a) What i s the probability that there are no major cracks in a 20 mile stretch of the hig hway? (b)What is the probability that the rst major crack occurs between 15 and 2 0 miles of the start of inspection? (c) Given that there are no cracks in the rst 5 miles inspected, what is the probability that there are no major cracks in th e next 15 miles? Solution. Let X denote the distance between major cracks. Then, X is an exponential 1 random variable with = E(X ) = 1 = 0.2 cracks/mile. 5 (a) 1 1 P (X > 20) = 20 0.2e0.2x dx = e0.2x 20 = e4 0.01831. (b) 20 P (15 < X < 20) = 15 0.2e0.2x dx = e0.2x 20 15 0.03154.
258 CONTINUOUS RANDOM VARIABLES (c) By the memoryless property, we have P (X > 15+5 X > 5) = P (X > 15) = 15 0.2e0.2x dx = e0.2x 15 0.04985
The exponential distribution is the only named continuous distribution that poss esses the memoryless property. To see this, suppose that X is memoryless continu ous random variable. Let g(x) = P (X > x). Since X is memoryless then P (X > t) = P (X > s+t X > s) = and this implies P (X > s + t) = P (X > s)P (X > t) Hence, g satis es the equation g(s + t) = g(s)g(t). Theorem 26.1 The only solution to th e functional equation g(s + t) = g(s)g(t) which is continuous from the right is g(x) = ex for some > 0. Proof. Let c = g(1). Then g(2) = g(1 + 1) = g(1)2 = c2 and g(3) = c3 so by simp e induction we can show that g(n) = cn for any positive in teger n. n 1 1 1 1 = g n g n g n = Now, let n be a positive integer, then g n 1 n 1 g n = c. Thus, g n = c n . 1 Next, let m and n be two positive integers. Then g m = g m n = n m m 1 1 1 1 g n + n + + n = g n = cn. Now, if t is a positive real number then we can nd a sequence tn of positive rational numbers such that limn tn = t. (This is known as the density property of the real numbers and is a topic d iscussed in a real analysis course). Since g(tn ) = ctn , the right continuity o f g implies g(t) = ct , t 0. inally, let = n c. Since 0 < c < 1, we have > 0. Moreover, c = e and therefore g(t) = et , t 0 It fo ows from the previous theorem t hat F (x) = P (X x) = 1 ex and hence f (x) = F (x) = ex which shows that X is expo tia y distributed. P (X > s + t) P (X > s + t and X > s) = P (X > s) P (X > s)
26 EXPONENTIAL RANDOM VARIABLES 259 Examp e 26.6 You ca the customer support department of a certain company and a re p aced on ho d. Suppose the amount of time unti a service agent assists you has an exponentia distribution with mean 5 minutes. Given that you have a ready been on ho d for 2 minutes, what is the probabi ity that you wi remain on ho d for a tota of more than 5 minutes? So ution. Let X represent the tota time o n ho d. Then X is an exponentia random 1 variab e with = 5 . Thus, P (X > 3 + 2 X > 2) = P (X > 3) = 1 F (3) = e 5 Example 26.7 Suppose that the duration of a p hone call (in minutes) is an exponential random variable X with = 0.1. (a) What is the probabi ity that a given phone ca asts more than 10 minutes? (b) Suppo se we know that a phone ca hast a ready asted 10 minutes. What is the probabi ity that it ast at east 10 more minutes? So ution. (a) P (X > 10) = 1 F (10) = e1 0.368 (b) P (X > 10 + 10 X > 10) = P (X > 10) 0.368 3
260 CONTINUOUS RANDOM VARIABLES Practice Problems Problem 26.1 Let X have an exponential distribution with a mean of 40. Compute P (X < 36). Problem 26.2 Let X be an exponential function with mean equals to 5. Graph f (x) and F (x). Problem 26.3 The lifetime (measured in years) of a radioa ctive element is a continuous random variable with the following p.d.f.: f (x) = x 1 100 e 100 0 t0 otherwise What is the robability that an atom of this element will decay within 50 years? Problem 26.4 The average number of radioactive articles assing through a coun ter during 1 millisecond in a lab ex eriment is 4. What is the robability that more than 2 milliseconds ass between articles? Problem 26.5 The life length of a com uter is ex onentially distributed with mean 5 years. You bought an old (w orking) com uter for 10 dollars. What is the robability that it would still wor k for more than 3 years? Problem 26.6 Su ose the wait time X for service at the ost o ce has an ex onential distribution with mean 3 minutes. A customer rst in l ine will be served immediately. (a) If you enter the ost o ce immediately behind another customer, what is the robability you wait over 5 minutes? (b) Under the same conditions, what is the robability of waiting between 2 and 4 minutes?
26 EXPONENTIAL RANDOM VARIABLES 261 Problem 26.7 During busy times buses arrive at a bus sto about every three minu tes, so if we measure x in minutes the rate of the exponentia distribution is = 1 . 3 (a) What is the probabi ity of having to wait 6 or more minutes for the b us? (b) What is the probabi ity of waiting between 4 and 7 minutes for a bus? (c ) What is the probabi ity of having to wait at east 9 more minutes for the bus given that you have a ready waited 3 minutes? Prob em 26.8 Ten years ago at a ce rtain insurance company, the size of c aims under homeowner insurance po icies h ad an exponentia distribution. Furthermore, 25% of c aims were ess than $1000. Today, the size of c aims sti has an exponentia distribution but, owing to i n ation, every c aim made today is twice the size of a simi ar c aim made 10 years ago. Determine the probabi ity that a c aim made today is ess than $1000. Prob em 26.9 The ifetime (in hours) of a ightbu b is an exponentia y distributed random variab e with parameter = 0.01. (a) What is the probabi ity that the igh t bu b is sti burning one week after it is insta ed? (b) Assume that the bu b was insta ed at noon today and assume that at 3:00pm tomorrow you notice that the bu b is sti working. What is the chance that the bu b wi burn out at som e time between 4:30pm and 6:00pm tomorrow? Prob em 26.10 The number of d ys th t el pse between the beginning of c lend r ye r nd the moment high risk driv er is involved in n ccident is exponenti lly distributed. An insur nce comp ny expects th t 30% of high risk drivers will be involved in n ccident during th e rst 50 d ys of c lend r ye r. Wh t portion of high risk drivers re expected to be involved in n ccident during the rst 80 d ys of c lend r ye r? Problem 26.11 The lifetime of printer costing 200 is exponenti lly distributed with me n 2 ye rs. The m nuf cturer grees to p y full refund to buyer if the print er f ils during the rst ye r following its purch se, nd one h lf refund if it
262 CONTINUOUS RANDOM VARIABLES f ils during the second ye r. If the m nuf cturer sells 100 printers, how much s hould it expect to p y in refunds? Problem 26.12 A device th t continuously me s ures nd records seismic ctivity is pl ced in remote region. The time, T, to f ilure of this device is exponenti lly distributed with me n 3 ye rs. Since the device will not be monitored during its rst two ye rs of service, the time to di scovery of its f ilure is X = m x (T, 2). Determine E[X]. Problem 26.13 A piece of equipment is being insured g inst e rly f ilure. The time from purch se unti l f ilure of the equipment is exponenti lly distributed with me n 10 ye rs. The insur nce will p y n mount x if the equipment f ils during the rst ye r, nd it will p y 0.5x if f ilure occurs during the second or third ye r. If f ilure occ urs fter the rst three ye rs, no p yment will be m de. At wh t level must x be s et if the expected p yment m de under this insur nce is to be 1000 ? Problem 26. 14 An insur nce policy reimburses dent l expense, X, up to m ximum bene t of 250 . The prob bility density function for X is: f (x) = ce0.004x x0 0 otherwise
where c is a constant. The time to f ilure of stribution with medi nent will work without xponenti l r ndom v ri X.
Calculate the median bene t for this olicy. Problem 26.15 component in n electronic device h s n exponenti l di n of four hours. C lcul te the prob bility th t the compo f iling for t le st ve hours. Problem 26.16 Let X be n e ble such th t P (X 2) = 2P (X > 4). Find the v ri nce of
26 EXPONENTIAL RANDOM VARIABLES 263 Problem 26.17 Customers rrive r ndomly nd independently t service window, nd the time between rriv ls h s n exponenti l distribution with me n of 12 m inutes. Let X equ l the number of rriv ls per hour. Wh t is P (X = 10)? Hint: W hen the time between successive rriv ls h s n exponenti l distri1 bution with me n (units of time), then the number of rriv ls per unit time h s Poisson di stribution with p r meter (me n) .
264 CONTINUOUS RANDOM VARIABLES 27 G mm nd Bet Distributions We st rt this section by introducing the G mm function de ned by () = 0 ey y 1 dy, > 0. For ex mple, (1) = 0 ey dy = ey = 1. 0 For > 1 we c n use integr tion by p rts with u = y 1 nd dv = ey dy to obt in () = ey y 1 + 0 0 ey ( 1)y 2 dy =( 1) 0 ey y 2 dy
=( 1)( 1) If n is positive integer gre ter th n 1 then by pplying the previous r el tion repe tedly we nd (n) =(n 1)(n 1) =(n 1)(n 2)(n 2) = =(n 1)(n 2 A amma random variable with arameters > 0 nd > 0 has a pdf f (x) = ex (x)1 () 0 if x 0 if x < 0 To see that f (t) is indeed a robability density function we have () = 0 ex x1 dx ex x1 dx () ey (y)1 dy () 1= 0 1= 0
27 GAMMA AND BETA DISTRIBUTIONS 265 where we used the substitution x = y. Note that the above computation invo ves a ( r) integral. Thus, the origin of the name of the random variable. Figure 27.1 sh ows some exam les of amma distributions. Figure 27.1 Theorem 27.1 If X is a amma random variable with arameters (, ) then ( ) E(X) = (b) V r(X) = 2 . So ution. (a) E(X) = 1 xex (x)1 dx () 0 1 = ex (x) dx () 0 ( + 1) = () =
266 (b) E(X 2 ) = CONTINUOUS RANDOM VARIABLES
1 x2 ex x1 dx () 0 1 x+1 ex dx = () 0 ( + 2) x+1 +2 ex = 2 dx
where the l st integr l is the integr l of the pdf of G mm r ndom v ri ble wi th p r meters ( + 2, ). Thus, E(X 2 ) = Fina y, V ar(X) = E(X 2 ) (E(X))2 = ( + 1) 2 = 2 2 ( + 1)( + 1) ( + 1) ( + 2) = = . 2 () 2 () 2 It is easy to see that when the parameter set is restricted to (, ) = (1, ) the gam ma distribution becomes the exponentia distribution. Another in1 teresting spec ia case is when the parameter set is (, ) = ( 2 , n ) where n is 2 a positive int eger. This distribution is ca ed the chi squared distribution. The gamma random variab e can be used to mode the waiting time unti a number of random events occurs. The number of random events sought is in the formul of f (x). Ex mple 2 7.1 In cert in city, the d ily consumption of electric power in millions of ki low tt hours c n be tre ted s r ndom v ri ble h ving g mm distribution wit h = 3 nd = 0.5. (a) What is the random variab e? What is the expected dai y con sumption? (b) If the power p ant of this city has a dai y capacity of 12 mi ion kWh, what is the probabi ity that this power supp y wi be inadequate on a giv en day? Set up the appropriate integra but do not eva uate.
27 GAMMA AND BETA DISTRIBUTIONS 267 So ution. (a) The random variab e is the dai y consumption of power in ki owatt hours. The expected dai y consumption is the expected va ue of a gamma distribut ed variab e with parameters = 3 nd = 1 which is E(X) = = 6. 2 x x 1 1 (b) The robabi ity is 23 (3) 12 x2 e 2 dx = 16 12 x2 e 2 dx
268 CONTINUOUS RANDOM VARIABLES Practice Problems Problem 27.1 Let X be a amma random variable with = 4 nd = P (2 < X < 4). Prob em 27.2 If X has a probabi ity density function given by f (x) = Find the mean and the variance. Prob em 27.3 Let X be a gamma random variab e with = 1.8 and = 3. Compute P (X > 3). Problem 27.4 A sherm n whose ver ge time for c tching sh is 35 minutes w nts to bring home ex ctly 3 shes. Wh t is the prob bility he wil l need between 1 nd 2 hours to c tch them? Problem 27.5 Suppose the time (in ho urs) t ken by professor to gr de complic ted ex m is r ndom v ri ble X h v ing g mm distribution with p r meters = 3 nd = 0.5. What is the probabi ity that it takes at most 1 hour to grade an exam? Prob em 27.6 Suppose the continuo us random variab e X has the fo owing pdf: f (x) = Find E(X 3 ). Prob em 27.7 L et X be the standard norma distribution. Show that X 2 is a gamma distribution with = = 1 . 2 1 2 x xe 2 16 1 . 2 Compute 4x2 e2x x>0 0 otherwise 0 if x > 0 otherwise
27 GAMMA AND BETA DISTRIBUTIONS 269 Prob em 27.8 Let X be a gamma random variab e with parameter (, ). Find E(etX ). P rob em 27.9 Show that the gamma density function with parameters > 1 nd > 0 1 h as a re ative maximum at x = ( 1). Problem 27.10 If comp ny employs n s lespers ons, its gross s les in thous nds of doll rs m y be reg rded s r ndom v ri ble h ving g mm distribution with = 80 n nd = 1 . If the sa es cost $8,000 per sa esperson, how many 2 sa esperson shou d the company emp oy to maximize the ex pected pro t? Prob em 27.11 Lifetimes of e ectrica components fo ow a Gamma dist ribution with = 3, 1 nd = 6 . (a) Give the density function(be very speci c to an y numbers), as we as the mean and standard deviation of the ifetimes. (b) The monetary va ue to the rm of a component with a ifetime of X is V = 3X 2 + X 1. Give the expected monetary va ue of a component.
270 CONTINUOUS RANDOM VARIABLES The Beta Distribution A random variab e is said to have a beta distribution with parameters a > 0 and b > 0 if its density is given by 1 xa1 (1 B(a,b) f (x) = x)b1 0 < x < 1 0 otherwise where 1 B(a, b) = 0 xa1 (1 x)b1 dx is the beta function. Examp e 27.2 Verify that the integra of the beta density function with paramete rs (2, 4) from to equa s 1. So ution. Using integration by parts, we have 1 f (x)dx = 20 0 1 1 x(1 x) dx = 20 x(1 x)4 (1 x)5 4 20 3 1 =1 0 Since the support of X is 0 < x < 1, the beta distribution is a popu ar probabi ity mode for proportions. The fo owing theorem provides a re ationship between the gamma and beta functions. Theorem 27.2 B(a, b) = (a)(b) . (a + b)
27 AMMA AND BETA DISTRIBUTIONS Proof. We have 1 271 (a + b)B(a, b) = 0 ta+b1 et dt 0 t xa1 (1 x)b1 dx ua1 1 u t b1 = 0 tb1 et dt 0 t du, u = xt = 0 et dt 0 ua1 (t u)b1 du = 0 u a1 du u et (t u)b1 dt = 0 ua1 eu du 0 ev v b1 dv, v = t u =(a)(b) Theorem 27.3 If X (a) E(X) = a+b (b) V ar(X) 1 1 xa (1 x)b1 dx E(X) = + 1) (a)(b) a(a)(a + b)
is a beta random variable with arameters a and b then a = (a+b)2ab . (a+b+1) Proof. (a) B(a, b) 0 B(a + 1, b) (a + 1)(b) (a + b) = = B(a, b) (a + = (a + b)(a + b)(a) a = a+b
(b) E(X 2 ) = 1 1 B(a + 2, b) (a + 2)(b) (a + b) = xa+1 (1 x)b1 dx = B(a, b) 0 B(a, b) (a + b + a)(b) a(a + 1)(a)(a + b) a(a + 1) = = (a + b)(a + b + 1)(a + b)(a) (a + b)(a + b + 1)
272 Hence, V ar(X) = E(X 2 )(E(X))2 = CONTINUOUS RANDOM VARIABLES a(a + 1) a2 ab = 2 2 (a + b + 1) (a + b)(a + b + 1) (a + b) (a + b) Exam le 27.3 The ro ortion of the gasoline su ly that is sold during the week is a beta random variable with a = 4 and b = 2. What is the robability of selli ng at least 90% of the stock in a given week? Solution. Since B(4, 2) = 3!1! 1 (4)(2) = = (6) 5! 20 the density function is f (x) = 20x3 (1 x). Thus, x4 x5 P (X > 0.9) = 20 x (1 x) dx = 20 4 5 0.9 3 1 1 0.08146 0.9 Exam le 27.4 Let f (x) = 12x2 (1 x) 0 x 1 0 otherwise Find the mean and the variance. Solution. Here, f (x) is a beta density function with a = 3 and b = 2. Thus, E(X) = a 1 = 3 and Var(X)= (a+b)2ab = 25 a+b 5 (a+b +1)
27 AMMA AND BETA DISTRIBUTIONS 273 Practice Problems Problem 27.12 Let f (x) = kx3 (1 x)2 0 x 1 0 otherwise (a) Find the value of k that makes f (x) a density function. (b) Find the mean a nd the variance of X. Problem 27.13 Su ose that X has a robability density fun ction given by f (x) = (a) Find F (x). (b) Find P (0.5 < X < 0.8). Problem 27.14 A management rm handles investment accounts for a large number of clients. The ercent of clients who tele hone the rm for information or services in a given mon th is a beta random variable with a = 4 and b = 3. (a) Find the density function f (x) of X. (b) Find the cumulative density function F (x). Problem 27.15 A com any markets a new roduct and surveys customers on their satisfaction with thei r roduct. The fraction of customers who are dissatis ed has a beta distribution w ith a = 2 and b = 4. What is the robability that no more than 30% of the custom ers are dissatis ed? Problem 27.16 Consider the following hy othetical situation. rade data indicates that on the average 27% of the students in senior engineeri ng classes have received A grades. There is variation among classes, however, an d the ro ortion X must be considered a random variable. From ast data we have measured a standard deviation of 15%. We would like to model the ro ortion X of A grades with a Beta distribution. (a) Find the density function of X. (b) Find the robability that more than 50% of the students had an A. 6x(1 x) 0 x 1 0 ot herwise
274 CONTINUOUS RANDOM VARIABLES Problem 27.17 Let X be a (1 x), 0 < x < 1 and 0 1). Problem 27.18 Let X at E(X n ) = B(a + n, b) beta random variable with density function f (x) = kx2 otherwise. (a) Find the value of k. (b) Find P (0.5 < X < be a beta distribution with arameters a and b. Show th . B(a, b)
Problem 27.19 Su ose that X is a beta distribution with arameters (a, b). Show that Y = 1 X is a beta distribution with arameters (b, a). Hint: Find FY (y) a nd then di erentiate with res ect to y. Problem 27.20 If the annual ro ortion of new restaurants that fail in a given city may be looked u on as a random variabl e having a beta distribution with a = 1 and b = 5, nd (a) the mean of this distri bution, that is, the annual ro ortion of new restaurants that can be ex ected t o fail in a given city; (b) the robability that at least 25 ercent of all new restaurants will fail in the given city in any one year. Problem 27.21 If the an nual ro ortion of erroneous income tax returns led with the IRS can be looked u on as a random variable having a beta distribution with arameters (2, 9), what is the robability that in any given year there will be fewer than 10 ercent er roneous returns? Problem 27.22 Su ose that the continuous random variable X has the density function f (x) = where B(a, b) = 0 1 xa1 (1 B(a,b) x)b1 0 < x < 1 0 otherwise 1 xa1 (1 x)b1 dx, a > 0, b > 0. If a = 5 and b = 6, what is E[(1 X)4 ]?
27 AMMA AND BETA DISTRIBUTIONS 275 Problem 27.23 The amount of im urity in batches of a chemical follow a Beta dist ribution with a = 1, and b = 2. (a) ive the density function, as well as the me an amount of im urity. (b) If the amount of im urity exceeds 0.7, the batch cann ot be sold. What ro ortion of batches can be sold?
276 CONTINUOUS RANDOM VARIABLES 28 The Distribution of a Function of a Random Variable Let X be a continuous random variable. Let g(x) be a function. Then g(X) is also a random variable. In this section we are interested in nding the robability de nsity function of g(X). The following exam le illustrates the method of nding the robability density function by nding rst its cdf. Exam le 28.1 If the robabilit y density of X is given by f (x) = 6x(1 x) 0 < x < 1 0 otherwise nd the robability density of Y = X 3 . Solution. We have F (y) =P (Y y) =P (X 3 y) =P (X y 3 ) y3 1 1 = 0 2 6x(1 x)dx =3y 3 2y Hence, f (y) = F (y) = 2(y 3 1), for 0 < y < 1 and 0 elsewhere Exam le 28.2 Let X be a random variable with robability density f (x). Find the robabi lity density function of Y = X . Solution. Clearly, F (y) = 0 for y 0. So assum e that y > 0. Then F (y) =P (Y y) =P ( X y) =P (y X y) =F (y) F (y) 1
28 THE DISTRIBUTION OF A FUNCTION OF A RANDOM VARIABLE277 Thus, f (y) = F (y) = f (y) + f (y) for y > 0 and 0 elsewhere The following theorem provides a formula for nding the probability density of g(X) for monotone g without the need for ndin g the distribution function. Theorem 28.1 Let X be a continuous random variable with pdf fX . Let g(x) be a monotone and di erentiable function of x. Suppose that g 1 (Y ) = X. Then the random variable Y = g(X) has a pdf given by fY (y) = fX [ g 1 (y)] Proof. Suppose rst that g() is increasing. Then FY (y) =P (Y y) = P (g(X) y) =P (X g 1 (y)) = FX (g 1 (y)) Di erentiating we nd fY (y) = dFY (y) d = fX [g 1 (y ] g 1 (y). dy dy d 1 g (y) . dy Now, suppose that g() is decreasing. Then FY (y) =P (Y y) = P (g(X) y) =P (X g 1 ( y)) = 1 X (g 1 (y)) Di erentiating we nd fY (y) = d d Y (y) = fX [g 1 (y)] g 1 (y) dy Exam le 28.3 Let X be a continuous random variable with df fX . ind the df of Y = X. Solution. By the revious theorem we have fY (y) = fX (y)
278 CONTINUOUS RANDOM VARIABLES Exam le 28.4 Let X be a continuous random variable with df fX . ind the df of Y = aX + b, a > 0. Solution. Let g(x) = ax + b. Then g 1 (y) = yb . a By the revious theorem, we have yb a 1 fY (y) = fX a Exam le 28.5 Su ose X is a random variable with the following density : f (x) = (a) ind the CD of X . (b) Find the pdf of X 2 . Solution. (a) X takes valu es in (0, ). Thus, F X (x) = 0 for x 0. Now, for x > 0 we have x (x2 1 , + 1) < x < .
X (x) = P ( X x) = x (x2 1 2 dx = tan1 x. + 1) Hence, 2
X (x) =
0 x0 tan1 x x > 0 for (b) X 2 also takes only ositive values, so the density fX 2 (x) = 0 x 0. 2 urthermore, X 2 (x) = P (X 2 x) = P ( X x) = tan1 x. So by di erentiating we get fX 2 (x) = 0 1 x(1+x) x0 x>0 Remark 28.1 In general, if a function does not have a unique inverse, we must su m over all ossible inverse values.
28 THE DISTRIBUTION O A UNCTION O A RANDOM VARIABLE279 Exam le 28.6 Let X be a continuous random variable with df fX . ind the df of Y = X 2 . Solution. L et g(x) = x2 . Then g 1 (y) = y. Thus, Y (y) = P (Y y) = P (X 2 y) = P ( y X ( y) X ( y). Di erentiate both sides to obtain, fX ( y) fX ( y) fY (y) = + 2 y
280 CONTINUOUS RANDOM VARIABLES Practice Problems Problem 28.1 Su ose fX (x) = (x) 1 e 2 2 2 and let Y = aX + b. ind fY (y). Problem 28.2 A rocess for re ning sugar yields u to 1 ton of ure sugar er day, but the actual amount roduced, is a random variable because of machine breakdo wns and other slowdowns. Su ose that X has the density function given by 2x 0 x 1 f (x) = 0 otherwise The com any is aid at the rate of $300 er ton for the r e ned sugar, but it also has a xed overhead cost of $100 er day. Thus, the daily ro t, in hundreds of dollars, is Y = 3X 1. ind robability density function for Y . Problem 28.3 Let X be a random variable with density function f (x) = 2x 0 x 1 0 otherwise ind the density function of Y = 8X 3 . Problem 28.4 Su ose X is an ex onential random variable with density function f (x) = ex x0 0 otherwise
What is the distribution of Y = eX ? Prob em 28.5 Gas mo ecu es move about with varying ve ocity which has, according to the Maxwe Bo tzmann aw, a probabi i ty density given by f (v) = cv 2 ev , 2 v0 The kinetic energy is given y Y = E = 1 mv 2 where m is the mass. What 2 is the density function of Y ?
0 x otherwise ind the robability density function of Y = cos X. Problem 28.8 Su ose X has t he uniform distribution on (0, 1). Com ute the robability density function and ex ected value of: (a) X , > 0 (b) ln X (c) eX (d) sin X Problem 28.9 An insur nc e comp ny sells n uto insur nce policy th t covers losses incurred by policy holder, subject to deductible of 100 . Losses incurred follow n exponenti l d istribution with me n 300. Wh t is the 95th percentile of ctu l losses th t exc eed the deductible? Problem 28.10 The time, T, th t m nuf cturing system is ou t of oper tion h s cumul tive distribution function F (t) = 1 0 2 2 t t>2 otherwise The resulting cost to the comp ny is Y = T 2 . Determine the density function of Y, for y > 4. Problem 28.11 An investment ccount e rns n nnu l interest r te R th t follows uniform distribution on the interv l (0.04, 0.08). The v lue o f 10,000 initi l investment in this ccount fter one ye r is given by V = 10, 000eR . Determine the cumul tive distribution function, FV (v) of V.
28 THE DISTRIBUTION OF a random varia le that ensity function of Y = ion over [, ]. That 1 2
A FUNCTION OF A RANDOM VARIABLE281 Pro lem 28.6 Let X e is uniformly distri uted in (0,1). Find the pro a ility d ln X. Pro lem 28.7 Let X e a uniformly distri uted funct is f (x) =
282 CONTINUOUS RANDOM VARIABLES Problem 28.12 An ctu ry models the lifetime of device using the r ndom v ri b le Y = 10X 0.8 , where X is n exponenti l r ndom v ri ble with me n 1 ye r. Det ermine the prob bility density function fY (y), for y > 0, of the r ndom v ri bl e Y. Problem 28.13 Let T denote the time in minutes for customer service repre sent tive to respond to 10 telephone inquiries. T is uniformly distributed on th e interv l with endpoints 8 minutes nd 12 minutes. Let R denote the ver ge r t e, in customers per minute, t which the represent tive responds to inquiries. F ind the density function fR (r) of R. Problem 28.14 The monthly pro t of Comp ny A c n be modeled by continuous r ndom v ri ble with density function fA . Comp ny B h s monthly pro t th t is twice th t of Comp ny A. Determine the prob bilit y density function of the monthly pro t of Comp ny B. Problem 28.15 Let X h ve nor m l distribution with me n 1 nd st nd rd devi tion 2. ( ) Find P ( X 1). (b) L et Y = eX . Find the probability density function fY (y) of Y. Problem 28.16 Let X be a uniformly distributed random variable on the interval (1, 1). 1 Show that Y = X 2 is a beta random variable with paramters ( 2 , 1). Problem 28.17 Let X be a random variable with density function f (x) = (a) Find the pdf of Y = 3X. ( b) Find the pdf of Z = 3 X. 3 2 x 2 0 1 x 1 otherwise
28 THE DISTRIBUTION OF A FUNCTION OF A RANDOM VARIABLE283 Problem 28.18 Let X be a continuous random variable with density function f (x) = 1 x 1 < x < 1 0 oth erwise Find the density 0 and Y = ln X, daily demand (Y, unction. f (x) = function of Y = X 2 . Problem 28.19 x2 If f (x) = xe 2 , for x > nd the density function for Y. Problem 28.20 A supplier faces a in proportion of her daily supply) with the following density f 2(1 x) 0 x 1 0 elsewhere
(a) Her pro t is given by Y = 10X 2. Find the density function of her daily pro t. ( b) What is her average daily pro t? (c) What is the probability she will lose mone y on a day?
Joint Distributions There are many situations which involve the presence of several random variables and we are interested in their joint behavior. For example: (i) A meteorologica l station may record the wind speed and direction, air pressure and the air temp erature. (ii) Your physician may record your height, weight, blood pressure, cho lesterol level and more. This chapter is concerned with the joint probability st ructure of two or more random variables de ned on the same sample space. 29 Jointly Distributed Random Variables Suppose that X and Y are two random variables de ned on the same sample space S. T he joint cumulative distribution function of X and Y is the function FXY (x, y) = P (X x, Y y) = P ({e S : X(e) x and Y (e) y}). Example 29.1 Consider the exper iment of throwing a fair coin and a fair die simultaneously. The sample space is S = {(H, 1), (H, 2), , (H, 6), (T, 1), (T, 2), , (T, 6)}. Let X be the number f head showing on the coin, X {0, 1}. Let Y be the number showing on the die, Y {1, 2, 3, 4, 5, 6}. Thus, if e = (H, 1) then X(e) = 1 and Y (e) = 1. Find FXY (1 , 2). 285
286 Solution. JOINT DISTRIBUTIONS FXY (1, 2) =P (X 1, Y 2) hat follows, individual cdfs e cdfs are obtained from the (X x) =P (X x, Y < ) =P y = lim P (X x, Y y) y = lim FXY (x, y) = FXY (x, ) y In a similar way, one can show that FY (y) = lim FXY (x, y) = FXY (, y). x =P ({(H, 1), (H, 2), (T, 1), (T, 2)}) 4 1 = = 12 3 In w will be referred to as marginal distributions. Thes joint cumulative distribution as follows FX (x) =P ( lim {X x, Y y})
It is easy to see that FXY (, ) = P (X < , Y < ) = 1. Also, FXY (, y) = 0. This follo s from 0 FXY (, y) = P (X < , Y y) P (X < ) = FX () = 0. Similarly, FXY (x,
29 JOINTLY DISTRIBUTED RANDOM VARIABLES 287 All joint probability statements about X and Y can be answered in terms of their joint distribution functions. For example, P (X > x, Y > y) =1 P ({X > x, Y > y }c ) =1 P ({X > x}c {Y > y}c ) =1 [P ({X x} {Y y}) =1 [P (X x) + P (Y y) Y y)] =1 FX (x) FY (y) + FXY (x, y) Also, if a1 < a2 and b1 < b2 then P (a1 < X a2 , b1 < Y b2 ) =P (X a2 , Y b2 ) P (X a2 , Y b1 ) P (X a1 , Y b2 ) + Y b1 ) =FXY (a2 , b2 ) FXY (a1 , b2 ) FXY (a2 , b1 ) + FXY (a1 , b1 ) This is c lear if you use the concept of area shown in Figure 29.1 Figure 29.1 If X and Y are both discrete random variables, we de ne the joint prob ability mass function of X and Y by pXY (x, y) = P (X = x, Y = y). The marginal probability mass function of X can be obtained from pXY (x, y) by pX (x) = P (X = x) = pXY (x, y). y:pXY (x,y)>0
288 JOINT DISTRIBUTIONS Similarly, we can obtain the marginal pmf of Y by pY (y) = P (Y = y) = x:pXY (x,y)>0 pXY (x, y). This simply means to nd the probability that X takes on a speci c value we sum acro ss the row associated with that value. To nd the probability that Y takes on a sp eci c value we sum the column associated with that value as illustrated in the nex t example. Example 29.2 A fair coin is tossed 4 times. Let the random variable X denote the number of heads in the rst 3 tosses, and let the random variable Y de note the number of heads in the last 3 tosses. (a) What is the joint pmf of X an d Y ? (b) What is the probability 2 or 3 heads appear in the rst 3 tosses and 1 o r 2 heads appear in the last three tosses? (c) What is the joint cdf of X and Y ? (d) What is the probability less than 3 heads occur in both the rst and last 3 tosses? (e) Find the probability that one head appears in the rst three tosses. S olution. (a) The joint pmf is given by the following table X\Y 0 1 2 3 pY (.) 0 1 2 3 1/16 1/16 0 0 1/16 3/16 2/16 0 0 2/16 3/16 1/16 0 0 1/16 1/16 2/16 6/16 6/ 16 2/16 pX (.) 2/16 6/16 6/16 2/16 1 (b) P ((X, Y ) {(2, 1), (2, 2), (3, 1), (3, 2)}) = p(2, 1) + p(2, 2) + p(3, 1) + p(3, 2) = 3 8 (c) The joint cdf is given by the following table X\Y 0 1 2 3 0 1 /16 2/16 2/16 2/16 1 2/16 6/16 8/16 8/16 2 2/16 8/16 13/16 14/16 3 2/16 8/16 14/ 16 1
29 JOINTLY DISTRIBUTED RANDOM VARIABLES 289 (d) P (X < 3, Y < 3) = F (2, 2) = 13 16 (e) P (X = 1) = P ((X, Y ) {(1, 0), (1, 1), (1, 2), (1, 3)}) = 1/16 + 3/16 + 2/16 = 3 8 Example 29.3 Suppose two balls are chosen from a box containing 3 white, 2 red a nd 5 blue balls. Let X = the number of white balls chosen and Y = the number of blue balls chosen. Find the joint pmf of X and Y. Solution. pXY (0, 0) =C(2, 2)/C(10, 2) = 1/45 pXY (0, 1) =C(2, 1)C(5, 1)/C(10, 2) = 10/45 pXY (0, 2) =C(5, 2)/C(10, 2) = 10/45 pXY (1, 0) =C(2, 1)C(3, 1)/C(10, 2) = 6/45 pXY (1, 1) =C(5, 1)C(3, 1)/C(10, 2) = 15/45 pXY (1, 2) =0 pXY (2, 0) =C(3, 2)/C( 10, 2) = 3/45 pXY (2, 1) =0 pXY (2, 2) =0 The pmf of X is 21 1 + 10 + 10 = 45 45 6 + 15 21 = 45 45 3 3 = 45 45 pX (0) =P (X = 0) = y:pXY (0,y)>0 pXY (0, y) = pXY (1, y) = y:pXY (1,y)>0 pX (1) =P (X = 1) = pX (2) =P (X = 2) = y:pXY (2,y)>0 pXY (2, y) =
290 The pmf of y is pY (0) =P (Y = 0) = x:pXY (x,0)>0 JOINT DISTRIBUTIONS pXY (x, 0) = pXY (x, 1) = x:pXY (x,1)>0 1+6+3 10 = 45 45 25 10 + 15 = 45 45 10 10 = 45 45 pY (1) =P (Y = 1) = pY (2) =P (Y = 2) = x:pXY (x,2)>0 pXY (x, 2) = Two random variables X and Y are said to be jointly continuous if there exists a function fXY (x, y) 0 with the property that for every su set C of R2 we have f XY (x, y)dxdy P ((X, Y ) C) = (x,y)C The function fXY (x, y) is called the joint pro a ility density function of X an d Y. If A and B are any sets of real num ers then y letting C = {(x, y) : x A, y B} we have P (X A, Y B) = B A fXY (x, y)dxdy As a result of this last equation we can write FXY (x, y) =P (X (, x], Y (, y]) y x = fXY (u, v)dudv It follows upon di erentiation that fXY (x, y) = 2 FXY (x, y) yx whenever the partial derivatives exist. Example 29.4 The cumulative distri ution function for the joint distri ution of the continuous random varia les X and Y is FXY (x, y) = 0.2(3x3 y + 2x2 y 2 ), 0 x 1, 0 y 1. Find fXY ( 1 , 1 ). 2 2
29 JOINTLY DISTRIBUTED RANDOM VARIABLES Solution. Since fXY (x, y) = we 1 , 1 ) = 2 2 17 20 291 2 FXY (x, y) = 0.2(9x2 + 8xy) yx
nd fXY (
Now, if X and Y are jointly continuous then they are individually continuous, an d their pro a ility density functions can e o tained as follows: P (X A) =P (X A, Y (, )) = A fXY (x, y)dydx fX (x, y)dx A = where fX (x) = fXY (x, y)dy is thus the pro a ility density function of X. Similarly, the pro a ility densit y function of Y is given y fY (y) = fXY (x, y)dx. Example 29.5 Let X and Y e random varia les with joint pdf fXY (x, y) = Determi ne (a) P (X 2 + Y 2 < 1), ( ) P (2X Y > 0), (c) P ( X + Y < 2). Solution. (a) 2 1 0 1 x, y 1 0 Otherwise 1 4 P (X 2 + Y 2 < 1) = 0 1 rdrd = . 4 4
292 (b) 1 1 y 2 JOINT DISTRIBUTIONS P (2X Y > 0) = 1 1 1 dxdy = . 4 2 Note that P (2X Y > 0) is the area of the region bounded by the lines y = 2x, x = 1, x = 1, y = 1 and y = 1. A gra h of this region will hel you understand the i ntegration rocess used above. (c) Since the s uare with vertices (1, 1), (1, 1), (1, 1), (1, 1) is com letely contained in the region 2 < x + y < 2, we have P ( X + Y < 2) = 1 Joint pdfs and joint cdfs for three or more random variables are o btained as straightforward generalizations of the above de nitions and conditions.
29 JOINTLY DISTRIBUTED RANDOM VARIABLES 293 Practice Problems Problem 29.1 A supermarket has two express lines. Let X and Y denote the number of customers in the rst and second line at any given time. The joint probability function of X and Y, pXY (x, y), is summarized by the following table X\Y 0 1 2 3 pY (.) 0 1 0.1 0.2 0.2 0.25 0 0.05 0 0 0.3 0.5 2 0 0.05 0.05 0.025 0.125 3 0 0 0.025 0.05 0.075 pX (.) 0.3 0.5 0.125 0.075 1 (a) Verify that pXY (x, y) is a joint probability mass function. (b) Find the pr obability that more than two customers are in line. (c) Find P ( X Y = 1), the probability that X and Y di er by exactly 1. (d) What is the marginal probability mass function of X? Problem 29.2 Two tire quality experts examine stacks of tir es and assign quality ratings to each tire on a 3 point scale. Let X denote the grade given by expert A and let Y denote the grade given by B. The following tab le gives the joint distribution for X, Y. X\Y 1 2 3 pY (.) Find P (X 2, Y 3). Pr o lem 29.3 In a randomly chosen lot of 1000 olts, let X e the num er that fail to meet a length speci cation, and let Y e the num er that fail to meet a diamet er speci cation. Assume that the joint pro a ility mass function of X and Y is giv en in the following ta le. 1 0.1 0.1 0.03 0.23 2 0.05 0.35 0.1 0.50 3 0.02 0.05 0.2 0.27 pX (.) 0.17 0.50 0.33 1
294 X\Y 0 1 2 pY (.) 0 1 0.4 0.12 0.15 0.08 0.1 0.03 0.65 0.23 2 0.08 0.03 0.01 0.12 JOINT DISTRIBUTIONS pX (.) 0.6 0.26 0.14 1 (a) Find P (X = 0, Y = 2). ( ) Find P (X > 0, Y 1). (c) P (X 1). (d) P (Y > 0). (e) Find the pro a ility that all olts in the lot meet the length speci cation. ( f) Find the pro a ility that all olts in the lot meet the diameter speci cation. (g) Find the pro a ility that all olts in the lot meet oth speci cation. Pro lem 29.4 Rectangular plastic covers for a compact disc (CD) tray have speci cations r egarding length and width. Let X and Y e the length and width respectively meas ured in millimeters. There are six possi le ordered pairs (X, Y ) and following is the corresponding pro a ility distri ution. X\Y 129 130 131 pY (.) (a) Find P (X = 130, Y = 15). ( ) Find P (X 130, Y 15). Pro lem 29.5 Suppose the random va ria les X and Y have a joint pdf fXY (x, y) = 0.0032(20 x y) 0 x, y 5 0 otherwis e 15 16 pX (.) 0.12 0.08 0.2 0.4 0.30 0.7 0.06 0.04 0.1 0.58 0.42 1 Find P (1 X 2, 2 Y 3). Pro lem 29.6 Assume the joint pdf of X and Y is fXY (x, y ) = xye 0 x2 +y 2 2 0 < x, y otherwise
29 JOINTLY DISTRIBUTED RANDOM VARIABLES (a) Find FXY (x, y). ( ) Find fX (x) and fY (y). 295 Pro lem 29.7 Show that the following function is not a joint pro a ility density function? fXY (x, y) = xa y 1a 0 x, y 1 0 otherwise where 0 < a < 1. What factor should you multiply fXY (x, y) to make it a joint p ro a ility density function? Pro lem 29.8 A device runs until either of two comp onents f ils, t which point the device stops running. The joint density functio n of the lifetimes of the two components, both me sured in hours, is fXY (x, y) = x+y 8 0 0 < x, y < 2 otherwise Wh t is the prob bility th t the device f ils during its rst hour of oper tion? P roblem 29.9 An insur nce comp ny insures l rge number of drivers. Let X be the r ndom v ri ble representing the comp nys losses under collision insur nce, nd let Y represent the comp nys losses under li bility insur nce. X nd Y h ve joint density function fXY (x, y) = 2x+2y 4 0 0 < x < 1, 0 < y < 2 otherwise
Wh t is the prob de lership sells er lso tries to Let X denote the
bility th t the tot l loss is t le st 1 ? Problem 29.10 A c r 0, 1, or 2 luxury c rs on ny d y. When selling c r, the de l persu de the customer to buy n extended w rr nty for the c r. number of luxury c rs sold in given d y, nd
296 JOINT DISTRIBUTIONS let Y denote the number of extended w rr nties sold. Given the following inform tion P (X = 0, Y = 0) = P (X = 1, Y P (X = 1, Y P (X = 2, Y P (X = 2, Y P (X = 2 , Y Wh t is the v ri nce of X? Problem 29.11 A comp ny is reviewing torn do d m ge cl ims under f rm insur nce policy. Let X be the portion of cl im represe nting d m ge to the house nd let Y be the portion of the s me cl im representin g d m ge to the rest of the property. The joint density function of X nd Y is f XY (x, y) = 6[1 (x + y)] x > 0, y > 0, x + y < 1 0 otherwise 1 6 1 = 0) = 12 1 = 1) = 6 1 = 0) = 12 1 = 1) = 3 1 = 2) = 6 Determine the prob bility th t the portion of cl im representing d m ge to the house is less th n 0.2. Problem 29.12 Let X nd Y be continuous r ndom v ri ble s with joint density function fXY (x, y) = 15y x2 y x 0 otherwise
Find the m rgin l density function of Y. Problem 29.13 Let X represent the f n insured utomobile involved in n ccident.
ge o
29 JOINTLY DISTRIBUTED RANDOM VARIABLES 297 Let Y represent the length of time the owner h s insured the utomobile t the t ime of the ccident. X nd Y h ve joint prob bility density function fXY (x, y) = 1 (10 64 xy 2 ) 2 x 10, 0 y 1 0 otherwise C lcul te the expected ge of n insured utomobile involved in n ccident. Pro blem 29.14 A device cont ins two circuits. The second circuit is b ckup for th e rst, so the second is used only when the rst h s f iled. The device f ils when nd only when the second circuit f ils. Let X nd Y be the times t which the rst nd second circuits f il, respectively. X nd Y h ve joint prob bility density f unction fXY (x, y) = 6ex e2y 0 < x < y < 0 otherwise Wh t is the expected time t which the device f ils? Problem 29.15 The future li fetimes (in months) of two components of m chine h ve the following joint dens ity function: fXY (x, y) = 6 (50 125000 x y) 0 < x < 50 y < 50 0 otherwise Wh t is the prob bility th t both components re still functioning 20 months fro m now? Problem 29.16 Suppose the r ndom v ri bles X nd Y h ve joint pdf fXY ( x, y) = Find P (X > Y ). x + y 0 x, y 1 0 otherwise Problem 29.17 Let X nd Y be r ndom losses with joint density function fXY (x, y ) = e(x+y) , x > 0, y > 0 nd 0 otherwise. An insur nce policy is written to reim burse X + Y. C lcul te the prob bility th t the reimbursement is less th n 1.
298 JOINT DISTRIBUTIONS
Problem 29.18 Let X nd Y be continuous r ndom v ri bles with joint cumul tive d istri1 bution FXY (x, y) = 250 (20xy x2 y xy 2 ) for 0 x 5 nd 0 y 5. Compute P (X > 2). Problem 29.19 Let X nd Y be continuous r ndom v ri bles with joint den sity function fXY (x, y) = Find P ( X Y X). 2 Problem 29.20 Let X nd Y be r ndo m v ri bles with common r nge {1, 2} nd such th t P (X = 1) = 0.7, P (X = 2) = 0.3, P (Y = 1) = 0.4, P (Y = 2) = 0.6, nd P (X = 1, Y = 1) = 0.2. ( ) Find the joint prob bility m ss function pXY (x, y). (b) Find the joint cumul tive distri bution function FXY (x, y) Problem 29.21 A device cont ins two components. The d evice f ils if either component f ils. The joint density function of the lifetim es of the components, me sured in hours, is f (s, t), where 0 < s < 1 nd 0 < t < 1. Wh t is the prob bility th t the device f ils during the rst h lf hour of op er tion? xy 0 x 2, 0 y 1 0 otherwise
30 INDEPENDENT RANDOM VARIABLES 299 30 Independent R ndom V ri bles Let X nd Y be two r ndom v ri bles de ned on the s me s mple sp ce S. We s y th t X nd Y re independent r ndom v ri bles if nd only if for ny two sets of re l numbers A nd B we h ve P (X A, Y B) = P (X A)P (Y B) Th t is the events E = { X A} nd F = {Y B} re independent. The following theorem expresses independence in terms of pdfs. Theorem 30.1 If X nd Y re discrete r ndom v ri bles, then X nd Y re independent if nd only if pXY (x, y) = pX (x)pY (y) where pX (x) nd pY (y) re the m rgin l pmfs of X nd Y respectively. Simil r result holds for continuous r ndom v ri bles where sums re repl ced by integr ls nd pmfs re re pl ced by pdfs. Proof. Suppose th t X nd Y re independent. Then by letting A = {x} nd B = {y} in Equ tion 30.1 we obt in P (X = x, Y = y) = P (X = x)P (Y = y ) th t is pXY (x, y) = pX (x)pY (y). Conversely, suppose th t pXY (x, y) = pX (x )pY (y). Let A nd B be ny sets of re l numbers. Then P (X A, Y B) = yB xA (30.1) pXY (x, y) pX (x)pY (y) yB xA = = yB pY (y) xA pX (x)
=P (Y B)P (X A) nt
nd thus equ tion 30.1 is s tis ed. Th t is, X nd Y
re independe
300 JOINT DISTRIBUTIONS Ex mple 30.1 1 A month of the ye r is chosen t r ndom (e ch with prob bility 12 ). Let X be the number of letters in the months n me, nd let Y be the number of d ys in the month (ignoring le p ye r). ( ) Write down the joint pdf of X nd Y . From this, compute the pdf of X nd the pdf of Y. (b) Find E(Y ). (c) Are the events X 6 nd Y = 30 independent? (d) Are X nd Y independent r ndom v ri bles? Sol ution. ( ) The joint pdf is given by the following t ble Y\ X 28 30 31 pX (x) 3 0 0 1 12 1 12 4 0 1 12 1 12 2 12 5 0 1 12 1 12 2 12 6 0 0 1 12 1 12 7 0 0 2 12 2 12 8 1 12 1 12 1 12 3 12 9 0 1 12 pY (y) 1 12 4 12 7 12 0 1 12 1
In the jointly continuous c se the condition of independence is equiv lent to fX Y (x, y) = fX (x)fY (y). It follows from the previous theorem, th t if you re g iven the joint pdf of the r ndom v ri bles X nd Y, you c n determine whether or not they re independent by c lcul ting the m rgin l pdfs of X nd Y nd determ ining whether or not the rel tionship fXY (x, y) = fX (x)fY (y) holds. Ex mple 3 0.2 The joint pdf of X nd Y is given by fXY (x, y) = Are X nd Y independent? 4 e2(x+y) 0 < x < , 0 < y < 0 Otherwise
1 4 7 (b) E(Y ) = 12 28 + 12 30 + 12 31 = 2 , P (Y = 30) = 12 = 1 , P (X 6, Y = 3 2 ) = P (X 6)P (Y = 30), the two 6 events re 8) = 0 = pX (5)pY (28) = 6 12 , X nd Y re
= 365 12 6 1 4 (c) We h ve P (X 6) = 12 30) = 12 = 1 . Since, P (X 6, Y = 30 independent. 1 1 (d) Since pXY (5, 2 dependent
30 INDEPENDENT RANDOM VARIABLES Solution. M rgin l density fX (x) is given by 301 fX (x) = 0 4e2(x+y) dy = 2e2x 0 2e2y dy = 2e2x , x > 0 Simil rly, the m rigin l density fY (y) is given by fY (y) = 0 4e2(x+y) dx = 2e2y 0 2e2x dx = 2e2y , y > 0 Now since fXY (x, y) = 4e2(x+y) = [2e2x ][2e2y ] = fX (x)fY (y) X nd Y re indepen dent Ex mple 30.3 The joint pdf of X nd Y is given by fXY (x, y) = 3(x + y) 0 x + y 1, 0 x, y < 0 Otherwise Are X nd Y independent? Solution. For the limit of integr tion see Figure 30.1 below. Figure 30.1 The m rgin l pdf of X is 1x fX (x) = 0 3 3(x + y)dy = 3xy + y 2 2 1x 0 3 = (1 x2 ), 0 x 1 2
302 The m rgin l pdf of Y is 1y JOINT DISTRIBUTIONS fY (y) = 0 3(x + y)dx = 3 2 x + 3xy 2 1y 0 3 = (1 y 2 ), 0 y 1 2 But 3 3 fXY (x, y) = 3(x + y) = (1 x2 ) (1 y 2 ) = fX (x)fY (y) 2 2 so th t X nd Y re dependent Ex mple 30.4 A m n nd wom n decide to meet t cert in loc tion. If e ch perso n independently rrives t time uniformly distributed between 2 nd 3 PM, nd th e prob bility th t the m n rrives t le st 10 minutes before the wom n. Solutio n. Let X represent the number of minutes p st 2 the m n rrives nd Y the number of minutes p st 2 the wom n rrives. X nd Y re independent r ndom v ri bles e ch uniformly distributed over (0,60). Then P (X + 10 < Y ) = = 10 fXY (x, y)dxdy x+10<y 2 60 y10 1 60 0 2 = x+10<y fX (x)fY (y)dxdy 1 60 dxdy 60 = (y 10)dy 0.347 10
The s me result holds for discrete r ndom v ri bles.
The following theorem provides necess ry nd su cient ri bles to be independent. Theorem 30.2 Two continuous re independent if nd only if their joint prob bility expressed s fXY (x, y) = h(x)g(y), < x < , < y
condition for two r ndom v r ndom v ri bles X nd Y density function c n be < .
30 INDEPENDENT RANDOM VARIABLES 303 Proof. Suppose rst th t X nd Y re independent. Then fXY (x, y) = fX (x)fY (y). Let h(x) = fX (x) nd g(y) = fY (y). Conversely, suppose th t fXY (x, y) = h(x)g (y). Let C = h(x)dx nd D = g(y)dy. Then CD = = h(x)dx g(y)dy h(x)g(y)dxdy = fXY (x, y)dxdy = 1 Furthermore, fX (x) = fXY (x, y)dy = h(x)g(y)dy = h(x)D nd fY (y) =
fXY (x, y)dx = h(x)g(y)dx = g(y)C. Hence, fX (x)fY (y) = h(x)g(y)CD = h(x)g(y) = fXY (x, y). This proves th t X nd Y re independent Ex mple 30.5 The joint pdf of X nd Y is given by fXY (x, y) = Are X nd Y independent? Solution. We h ve fXY (x, y) = xye (x2 +y 2 ) 2 xye (x2 +y 2 ) 2 0 0 x, y < Otherwise = xe 2 ye 2
x2 y2
By the previous theorem, X
nd Y re independent
304 Ex mple 30.6 The joint pdf of X nd Y is given by fXY (x, y) = Are X nd Y i ndependent? Solution. Let I(x, y) = Then JOINT DISTRIBUTIONS x + y 0 x, y < 1 0 Otherwise 1 0 x < 1, 0 y < 1 0 otherwise
fXY (x, y) = (x + y)I(x, y) which cle rly does not f ctor into p rt depending only on x nd nother depending only on y. Thus, by the previous theorem X nd Y re dependent Ex mple 30.7 Let X nd Y be two independent r ndom v ri ble with X h ving norm l distribution with me n nd v ri nce 1 nd Y being the st nd rd norm l distribution. ( ) Find the density of Z = min{X, Y }. (b) For e ch t R c lcul te P (m x(X, Y ) min(X, Y ) > t). Solution. ( ) Fix re l number z. Then FZ (z) =P (Z z) = 1 P (min(X, Y ) > z) =1 P (X > z)P (Y > z) = 1 (1 (z ))(1 nce, fZ (z) = (1 (z ))(z) + (1 (z))(z ). (b) I t 0 then P (max(X, Y ) mi = 1. I t > 0 then P (max(X, Y ) min(X, Y ) > t) =P ( X Y > t) t + =1 2 Not t X Y is normal with mean and variance 2 t 2
30 INDEPENDENT RANDOM VARIABLES 305
Example 30.8 Suppose X1 , , Xn are independent and identically distributed rando m variables with cdf FX (x). De ne U and L as U =max{X1 , X2 , , Xn } L =min{X1 , X2 , , Xn } (a) Find the cdf of U. (b) Find the cdf of L. (c) Are U and L indepe ndent? Solution. (a) First note the following equivalence of events {U u} {X1 u, X2 u, , Xn u}. Thus, FU (u) =P (U u) = P (X1 u, X2 u, , Xn u) =P ( P (Xn u) = (F (X))n (b) Note the following equivalence of events {L > l} {X1 > l, X2 > l, , Xn > l}. Thus, FL (l) =P (L l) = 1 P (L > l) =1 P (X1 > l, X2 > l, , Xn > l) =1 P (X1 > l)P (X2 > l) P (Xn > l) =1 (1 F (X))n (c) No. First note t at P (L > l) = 1 F )L(l). From the de nition of cdf there must be a number l0 such that FL (l0 ) = 1. Thus, P (L > l0 ) = 0. But P (L > l0 U u) = 0 for any u < l 0 . This shows that P (L > l0 U u) = P (L > l0 )
306 JOINT DISTRIBUTIONS Practice Problems Problem 30.1 Let X and Y be random variables = (a) Are X and Y independent? (b) Find P (X 30.2 The random vector (X, Y ) is said to be R in the plane if, for some constant c, its R 0 otherwise e(x+y) 0 x, y 0 otherwise with joint pdf given by fXY (x, y) < Y ). (c) Find P (X < a). Problem uniformly distributed over a region joint pdf is fXY (x, y) = c (x, y)
1 (a) Show that c = A(R) where A(R) is the area of the region R. (b) Suppose tha t R = {(x, y) : 1 x 1, 1 y 1}. Show that X and Y are independent, with each being distributed uniformly over (1, 1). (c) Find P (X 2 + Y 2 1). Problem 30.3 Let X and Y be random variables with joint pdf given by fXY (x, y) = 1 (a) Find P (X 3 , Y 2 ). 4 (b) ind X (x) and Y (y). (c) Are X and Y indepen dent? 6(1 y) 0 x y 1 0 otherwise Problem 30.4 Let X and Y have the joint pd given by XY (x, y) = (a) ind k. (b ) ind X (x) and Y (y). (c) Are X and Y independent? kxy 0 x, y 1 0 otherwise

30 INDEPENDENT RANDOM VARIABLES Problem 30.5 Let X and Y have joint density XY (x, y) = kxy 2 0 x, y 1 0 otherwise 307 (a) ind k. (b) Compute the marginal densities o X and o Y . (c) Compute P (Y > 2X). (d) Compute P ( X Y < 0.5). (e) Are X and Y independent? Problem 30.6 S uppose the joint density of random variables X and Y is given by fXY (x, y) = (a ) Find k. (b) Are X and Y independent? (c) Find P (X > Y ). Problem 30.7 Let X a nd Y be continuous random variables, with the joint probability density function fXY (x, y) = (a) Find fX (x) and fY (y). (b) Are X and Y independent? (c) Find P (X + 2Y < 3). Problem 30.8 Let X and Y have joint density fXY (x, y) = x y 3 x , 0 x 0 otherwise 4 9 3x2 +2y 24 kx2 y 3 1 x, y 2 0 otherwise 0 0 x, y 2 otherwise (a) Compute the marginal densities of X and Y. (b) Compute P (Y > 2X). (c) Are X and Y independent?
308 JOINT DISTRIBUTIONS Problem 30.9 A study is being conducted in which the health of two independent g roups of ten policyholders is being monitored over a one year period of time. In dividual participants in the study drop out before the end of the study with pro bability 0.2 (independently of the other participants). What is the probability that at least 9 participants complete the study in one of the two groups, but no t in both groups? Problem 30.10 The waiting time for the rst claim from a good dr iver and the waiting time for the rst claim from a bad driver are independent and follow exponential distributions with means 6 years and 3 years, respectively. What is the probability that the rst claim from a good driver will be led within 3 years and the rst claim from a bad driver will be led within 2 years? Problem 30. 11 An insurance company sells two types of auto insurance policies: Basic and De luxe. The time until the next Basic Policy claim is an exponential random variab le with mean two days. The time until the next Deluxe Policy claim is an indepen dent exponential random variable with mean three days. What is the probability t hat the next claim will be a Deluxe Policy claim? Problem 30.12 Two insurers pro vide bids on an insurance policy to a large company. The bids must be between 20 00 and 2200 . The company decides to accept the lower bid if the two bids di er by 20 or more. Otherwise, the company will consider the two bids further. Assume t hat the two bids are independent and are both uniformly distributed on the inter val from 2000 to 2200. Determine the probability that the company considers the two bids further. Problem 30.13 A family buys two policies from the same insuran ce company. Losses under the two policies are independent and have continuous un iform distributions on the interval from 0 to 10. One policy has a deductible of 1 and the other has a deductible of 2. The family experiences exactly one loss under each policy. Calculate the probability that the total bene t paid to the fam ily does not exceed 5.
30 INDEPENDENT RANDOM VARIABLES 309 Problem 30.14 In a small metropolitan area, annual losses due to storm, re, and t heft are assumed to be independent, exponentially distributed random variables w ith respective means 1.0, 1.5, and 2.4 . Determine the probability that the maxi mum of these losses exceeds 3. Problem 30.15 A device containing two key compone nts fails when, and only when, both components fail. The lifetimes, X and Y , of these components are independent with common density function f (t) = et , t > 0 . The cost, Z, of operating the device until failure is 2X + Y. Find the probabi lity density function of Z. Problem 30.16 A company o ers earthquake insurance. An nual premiums are modeled by an exponential random variable with mean 2. Annual claims are modeled by an exponential random variable with mean 1. Premiums and c laims are independent. Let X denote the ratio of claims to premiums. What is the density function of X? Problem 30.17 Let X and Y be independent continuous rand om variables with common density function 1 0<x<1 fX (x) = fY (x) = 0 otherwise What is P (X 2 Y 3 )? Problem 30.18 Suppose that discrete random variables X and Y each take only the values 0 and 1. It is known that P (X = 0 Y = 1) = 0.6 and P (X = 1 Y = 0) = 0.7. Is it possible that X and Y are independent? Justify you r conclusion. Problem 30.19 Let X and Y be two discrete random variables with jo int distribution given by the following table. Y\ X 2 4 1 1 + 2 1 + 22 5 1 + 22 1 +
We assume that 0.25 1 2.5 and 0 2 0.35. ind 1 and 2 when X and Y are independ oblem 30.20 An insur nce policy is written to cover loss X where X h s density function fX (x) = 3 2 x 8 0 0x2 otherwise The time T (in hours) to process cl im of size x, where 0 x 2, is uniformly di stributed on the interv l from x to 2x. C lcul te the prob bility th t r ndoml y chosen cl im on this policy is processed in three hours or more.
31 SUM OF TWO INDEPENDENT RANDOM VARIABLES 311 31 Sum of Two Independent R ndom V ri bles In this section we turn to the import nt question of determining the distributio n of sum of independent r ndom v ri bles in terms of the distributions of the individu l constituents. 31.1 Discrete C se In this subsection we consider only sums of discrete r ndom v ri bles, reserving the c se of continuous r ndom v ri bles for the next subsection. We consider he re only discrete r ndom v ri bles whose v lues re nonneg tive integers. Their d istribution m ss functions re then de ned on these integers. Suppose X nd Y re two independent discrete r ndom v ri bles with pmf pX (x) nd pY (y) respectivel y. We would like to determine the pmf of the r ndom v ri ble X + Y. To do this, we note rst th t for ny nonneg tive integer n we h ve n {X + Y = n} = k=0 Ak where Ak = {X = k} {Y = n k}. Note th t Ai Aj = for i = j. Since the Ai s re p i rwise disjoint nd X nd Y re independent, we h ve n P (X + Y = n) = k=0 P (X = k)P (Y = n k). Thus, pX+Y (n) = pX (n) pY (n) where pX (n) pY (n) is c lled the convolution of pX nd pY . Ex mple 31.1 A die is rolled twice. Let X nd Y be the outcomes, nd let Z = X + Y be the sum of these outcomes. Find the prob bility m ss function of Z. Solution. Note th t X nd Y h ve the common pmf : x pX 1 2 3 4 5 1/6 1/6 1 /6 1/6 1/6 6 1/6
312 JOINT DISTRIBUTIONS The prob bility m ss function of Z is then the convolution of pX with itself. Th us, 1 P (Z = 2) =pX (1)pX (1) = 36 2 P (Z = 3) =pX (1)pX (2) + pX (2)pX (1) = 36 3 P (Z = 4) =pX (1)pX (3) + pX (2)pX (2) + pX (3)pX (1) = 36 Continuing in this w y we would nd P (Z = 5) = 4/36, P (Z = 6) = 5/36, P (Z = 7) = 6/36, P (Z = 8) = 5/36, P (Z = 9) = 4/36, P (Z = 10) = 3/36, P (Z = 11) = 2/36, nd P (Z = 12) = 1/36 Ex mple 31.2 Let X nd Y be two independent Poisson r ndom v ri bles with respective p r meters 1 and 2 . Compute the pmf of X + Y. So ution. For every posi tive integer we have n {X + Y = n} = k=0 Ak where Ak = {X = k, Y = n k} for 0 k n. Moreover, Ai Aj = for i = j. Thus, n pX+Y (n) =P (X + Y = n) = k=0 n P (X = k, Y = n k) = k=0 n P (X = k)P (Y = n k) e1 k=0 (1 +2 ) k=0 n = =e = = k 2 nk 2 1 e k! (n k)! n nk k 2 1 k!(n k)! e e (1 +2 ) n! (1 +2 ) k=0 n! k nk k!(n k)! 1 2 n! (1 + 2 )n
31 SUM OF TWO INDEPENDENT RANDOM VARIABLES Thus, X + Y is a Poisson random varia b e with parameters 1 + 2 313 Examp e 31.3 Let X and Y be two independent binomia random variab es with respe ctive parameters (n, p) and (m, p). Compute the pmf of X + Y. So ution. X repres ents the number of successes in n independent tria s, each of which resu ts in a success with probabi ity p; simi ar y, Y represents the number of successes in m independent tria s, each of which resu ts in a success with probabi ity p. Hen ce, as X and Y are assumed to be independent, it fo ows that X + Y represents t he number of successes in n + m independent tria s, each of which resu ts in a s uccess with probabi ity p. So X + Y is a binomia random variab e with parameter s (n + m, p) Examp e 31.4 A ice and Bob ip bias coins independent y. A ices coin c omes up heads with probabi ity 1/4, whi e Bobs coin comes up head with probabi it y 3/4. Each stop as soon as they get a head; that is, A ice stops when she gets a head whi e Bob stops when he gets a head. What is the pmf of the tota amount of ips unti both stop? (That is, what is the pmf of the combined tota amount of ips for both A ice and Bob unti they stop?) So ution. Let X and Y be the number of ips unti A ice and Bob stop, respective y. Thus, X+Y is the tota number of i ps unti both stop. The random variab es X and Y are independent geometric rando m variab es with parameters 1/4 and 3/4, respective y. By convo ution, we have n1 pX+Y (n) = = 1 4n k=1 n1 1 4 3 4 k1 3 4 1 4 nk1 3k = k=1 3 3n1 1 2 4n
314 JOINT DISTRIBUTIONS Practice Prob ems Prob em 31.1 Let X and Y be two independent discrete random variab es with proba bi ity mass functions de ned in the tab es be ow. Find the probabi ity mass functi on of Z = X + Y. x 0 1 2 3 pX (x) 0.10 0.20 0.30 0.40 y 0 1 2 pY (y) 0.25 0.40 0 .35 Prob em 31.2 Suppose X and Y are two independent binomia random variab es with respective parameters (20, 0.2) and (10, 0.2). Find the pmf of X + Y. Prob em 31 .3 Let X and Y be independent random variab es each geometrica y distributed wi th paramter p, i.e. pX (n) = pY (n) = p(1 p)n1 n = 1, 2, 0 otherwise Find the probability mass function of X + Y. Problem 31.4 Consider the following two experiments: the rst has outcome X taking on the values 0, 1, and 2 with equ al probabilities; the second results in an (independent) outcome Y taking on the value 3 with probability 1/4 and 4 with probability 3/4. Find the probability m ass function of X + Y. Problem 31.5 An insurance company determines that N, the number of claims received 1 in a week, is a random variable with P [N = n] = 2n+ 1 , where n 0. The company also determines that the number o claims received in a given week is independent o the number o claims received in any other week. Determine the probability that exactly seven claims will be received during a g iven two week period. Problem 31.6 Suppose X and Y are independent, each having Poisson distribution with means 2 and 3, respectively. Let Z = X + Y. ind P (X + Y = 1).
31 SUM O TWO INDEPENDENT RANDOM VARIABLES 315 Problem 31.7 Suppose that X has Poisson distribution with parameter and that Y h as geometric distribution with parameter p and is independent of X. Find simp e formu as in terms of and p for the fo owing probabi ities. (The formu as shou d not invo ve an in nite sum.) (a) P (X + Y = 2) (b) P (Y > X) Prob em 31.8 An insu rance company has two c ients. The random variab es representing the c aims ed b y each c ient are X and Y. X and Y are indepedent with common pmf x pX (x) 0 1 2 y 0.5 0.25 0.25 pY (y) Find the probabi ity density function of X + Y. Prob em 31.9 Let X and Y be two independent random variab es with pmfs given by pX (x) = x = 1, 2, 3 0 otherwise 1 2 1 3 1 6 1 3 y=0 y=1 pY (y) = y=2 0 otherwise Find the probabi ity density function of X + Y. Prob em 31.10 Let X and Y be two independent identica y distributed geometric distributions with parameter p. Show that X + Y is a negative binomia distribut ion with parameters (2, p). Prob em 31.11 Let X, Y, Z be independent Poisson ran dom variab es with E(X) = 3, E(Y ) = 1, and E(Z) = 4. What is P (X + Y + Z 1)? P rob em 31.12 If the number of typographica errors per page type by a certain ty pist fo ows a Poisson distribution with a mean of , nd the probabi ity that the t ota number of errors in 10 random y se ected pages is 10.
316 JOINT DISTRIBUTIONS 31.2 Continuous Case In this subsection we consider the continuous version of the prob em posed in Se ction 31.1: How are sums of independent continuous random variab es distributed? Examp e 31.5 Let X and Y be two random variab es with joint probabi ity density fXY (x, y) = 6e3x2y x > 0, y > 0 0 e sewhere Find the probabi ity density of Z = X + Y. So ution. Integrating the joint proba bi ity density over the shaded region of Figure 31.1, we get a ay FZ (a) = P (Z a) = 0 0 6e3x2y dxdy = 1 + 2e3a 3e2a and di erentiating with respect to a we here nd fZ (a) = 6(e2a e3a ) for a > 0 and 0 e sew
Figure 31.1 The above process can be genera ized with the use of convo utions wh ich we de ne next. Let X and Y be two continuous random variab es with
31 SUM OF TWO INDEPENDENT RANDOM VARIABLES 317 probabi ity density functions fX (x) and fY (y), respective y. Assume that both fX (x) and fY (y) are de ned for a rea numbers. Then the convo ution fX fY of f X and fY is the function given by (fX fY )(a) = fX (a y)fY (y)dy fY (a x)fX (x)dx = This de nition is ana ogous to the de nition, given for the discrete case, of the co nvo ution of two probabi ity mass functions. Thus it shou d not be surprising th at if X and Y are independent, then the probabi ity density function of their su m is the convo ution of their densities. Theorem 31.1 Let X and Y be two indepen dent random variab es with density functions fX (x) and fY (y) de ned for a x an d y. Then the sum X + Y is a random variab e with density function fX+Y (a), whe re fX+Y is the convo ution of fX and fY . Proof. The cumu ative distribution fun ction is obtained as fo ows: FX+Y (a) =P (X + Y a) = x+ya ay ay fX (x)fY (y)dxdy fX (x)dxfY (y)dy = fX (x)fY (y)dxdy = FX (a y)fY (y)dy = Di erentiating the previous equation with respect to a we nd fX+Y (a) = = d da FX (a y)fY (y)dy d FX (a y)fY (y)dy da fX (a y)fY (y)dy = =(fX fY )(a)
318 JOINT DISTRIBUTIONS Examp e 31.6 Let X and Y be two independent random variab es uniform y distribut ed on [0, 1]. Compute the distribution of X + Y. So ution. Since fX (a) = fY (a) = by the previous theorem 1 1 0a1 0 otherwise fX+Y (a) = 0 fX (a y)dy.
Now the integrand is 0 un ess 0 a y 1(i.e. un ess a 1 y a) and then it is 1. So if 0 a 1 then a fX+Y (a) = 0 dy = a. 1 If 1 < a < 2 then fX+Y (a) = dy = 2 a. a1 Hence, fX+Y (a) = a 0a1 2a 1<a<2 0 otherwise
Examp e 31.7 Let X and Y be two independent exponentia random variab es with co mmon parameter . Compute fX+Y (a). So ution. We have fX (a) = fY (a) = If a 0 the n ea 0a 0 otherwise fX+Y (a) = fX (a y)fY (y)dy a =2 0 ea dy = a2 ea
31 SUM OF TWO INDEPENDENT RANDOM VARIABLES If a < 0 then fX+Y (a) = 0. Hence, fX +Y (a) = a2 ea 0a 0 otherwise 319 Examp e 31.8 Let X and Y be two independent random variab es, each with the stan dard norma density. Compute fX+Y (a). So ution. We have x2 1 fX (a) = fY (a) = e 2 . 2 By Theorem 31.1 we have (ay)2 y2 1 fX+Y (a) = e 2 e 2 dy 2 1 a2 (y a )2 2 e = e 4 dy 2 a2 1 w=y a 2 The ex ression in the brackets equals 1, since it is the integral of the normal 1 density function with = 0 and = 2 . Hence, a2 1 fX+Y (a) = e 4 4 Exam le 31.9 Let X and Y be two inde endent gamma random variables with res ecti ve arameters (s, ) and (t, ). Show that X + Y is a gamma random variab e with par ameters (s + t, ). So ution. We have fX (a) = ea (a)s1 (s)
ea (a)t1 (t)
and
Y (a) =
320 JOINT DISTRIBUTIONS By Theorem 31.1 we have a 1 X+Y (a) = e(ay) [(a y)]s1 ey (y)t1 dy (s)(t) 0 t1 dy (s)(t) 0 s+t ea as+t1 1 y = (1 x)s1 xt1 dx, x = (s)(t) a 0 But we have 9.8 that 1 (1 x)s1 xt1 dx = B(s, t) = 0 (s)(t) . (s + t) Thus, X+Y (a) = ea (a)s+t1 (s + t) Example 31.10 The percentages o copper and iron in a certain kind o ore are, r espectively, X and Y. I the joint density o these two random variables is give n by XY (x, y) = 3 (5x 11 0 + y) x, y > 0, x + 2y < 2 elsewhere use the distribution unction techni ue to nd the probability density o Z = X + Y. Solution. Note rst that the region o integration is the interior o the trian gle with vertices at (0, 0), (0, 1), and (2, 0). rom the gure we see that (a) = 0 i a < 0. Now, the two lines x + y = a and x + 2y = 2 intersect at (2a 2, 2 a). I 0 a < 1 then a ay 0 Z (a) = P (Z a) = 0 3 3 (5x + y)dxdy = a3 11 11 I 1 a < 2 then Z (a) =P (Z a) 2a = 0 3 = 11 3 (5x + y)dxdy + 11 0 7 7 a3 + 9a2 8a + 3 3 ay 1 2a 0 22y 3 (5x + y)dxdy 11

31 SUM O TWO INDEPENDENT RANDOM VARIABLES I a 2 then Z (a) = 1. Di erentiating with respect to a we nd 9 2 a 0<a1 11 3 2 (7a + 18a 8) 1 < a < 2 Z (a) = where 321
11 0 e
322 JOINT DISTRIBUTIONS Practice Problems Problem 31.13 Let X be an exponential random variable with parameter and Y be an exponentia random variab e with parameter 2 independent of X. Find the probabi ity density function of X + Y. Prob em 31.14 Let X be an exponentia random vari ab e with parameter and Y be a uniform random variab e on [0,1] independent of X . Find the probabi ity density function of X + Y. Prob em 31.15 Let X and Y be t wo independent random variab es with probabi ity density functions (p.d.f.) , fX and fY respective y. Find the pdf of X + 2Y. Prob em 31.16 Consider two indepen dent random variab es X and Y. Let fX (x) = 1 x 2 if 0 x 2 and 0 otherwise. Let fY (y) = 2 2y for 0 y 1 and 0 otherwise. Find the probabi ity density function o f X + Y. Prob em 31.17 Let X and Y be two independent and identica y distribute d random variab es with common density function f (x) = 2x 0 < x < 1 0 otherwise Find the probabi ity density function of X + Y. Prob em 31.18 Let X and Y be ind ependent exponentia random variab es with pairwise distinct respective paramete rs nd . Find the pro a ility density function of X + Y. Pro lem 31.19 A device c ontaining two key components fails when and only when oth components fail. The lifetime, T1 and T2 , of these components are independent with a common density function given y fT1 (t) = fT2 (t) = et t>0 0 otherwise
31 SUM OF TWO INDEPENDENT RANDOM VARIABLES 323 The cost, X, of operating the device until failure is 2T1 +T 2. Find the density function of X. Pro lem 31.20 Let X and Y e independent random varia les with d ensity functions fX (x) = 1 2 0 1 2 1 x 1 otherwise fY (y) = 3y5 0 otherwise Find the pro a ility density function of X + Y. Pro lem 31.21 Let X and Y e ind ependent random varia les with density functions fX (x) = 0<x<2 0 otherwise 0<y< 2 0 otherwise y 2 1 2 fY (y) = Find the pro a ility density function of X + Y. Pro lem 31.22 Let X and Y e ind ependent random varia les with density functions (x1 ) 1 2 fX (x) = e 21 21 2 and (y2 ) 1 2 fY (y) = e 22 22 2 Find the probability den ity function of X + Y. Problem 31.23 Let X have a unifo rm di tribution on the interval (1, 3). What i the probability that the um of 2 independent ob ervation of X i greater than 5?
324 JOINT DISTRIBUTIONS Problem 31.24 The life (in day ) of a certain machine ha an exponential di trib ution with a mean of 1 day. The machine come upplied with one pare. Find the den ity function (t mea ure in day ) of the combined life of the machine and it pare if the life of the pare ha the ame di tribution a the r t machine, but i independent of the r t machine. Problem 31.25 X1 and X2 are independent expon ential random variable each with a mean of 1. Find P (X1 + X2 < 1).
32 CONDITIONAL DISTRIBUTIONS: DISCRETE CASE 325 32 Conditional Di tribution : Di crete Ca e Recall that for any two event E and F the conditional probability of E given F i de ned by P (E F ) P (E F ) = P (F ) provided that P (F ) > 0. In a similar way , if X and Y are discrete random variables then we de ne the conditional probabili ty mass function of X given that Y = y by pX Y (x y) =P (X = x Y = y) P (X = x, Y = y) = P (Y = y) pXY (x, y) = pY (y) provided that pY (y) > 0. Example 32.1 Su ppose you and me are tossing two fair coins independently, and we will stop as s oon as each one of us gets a head. (a) Find the chance that we stop simultaneous ly. (b) Find the conditional distribution of the number of coin tosses given tha t we stop simultaneously. Solution. (a) Let X be the number of times I have to t oss my coin before getting a head, and Y be the number of times you have to toss your coin before getting a head. So X and Y are independent identically distrib uted geometric random 1 variables with parameter p = 2 . Thus, (32.1) P (X = Y ) = k=1 P (X = k, Y = k) = k=1 P (X = k)P (Y = k) = k=1 1 1 = k 4 3
326 JOINT DISTRIBUTIONS (b) Notice that given the event [X = Y ] the number of coin tosses is well de ned and it is X (or Y ). So for any k 1 we have P (X = k, Y = k) = P (X = k Y = k) = P (X = Y ) 1 4k 1 3 3 = 4 1 4 k1 . Thus given [X = Y ], the number of tosses follows a geometric distribution 3 wit h parameter p = 4 Sometimes it is not the joint distribution that is known, but rather, for each y, one knows the conditional distribution of X given Y = y. If one also knows the distribution of Y, then one can recover the joint distributio n using (32.1). We also mention one more use of (32.1): pX (x) = y pXY (x, y) pX Y (x y)pY (y) y = (32.2) Thus, given the conditional distribution of X given Y = y for each possible valu e y, and the (marginal) distribution of Y, one can compute the (marginal) distri bution of X, using (32.2). The conditional cumulative distribution of X given th at Y = y is de ned by FX Y (x y) =P (X x Y = y) = ax pX Y (a y) Note that if conditional follows from Y (x y) = pX X and Y are independent, then the conditional mass function and the distribution function are the same as the unconditional ones. This the next theorem. Theorem 32.1 If X and Y are independent, then pX (x).
32 CONDITIONAL DISTRIBUTIONS: DISCRETE CASE Proof. We have pX Y (x y) =P (X = x Y = y) P (X = x, Y = y) = P (Y = y) P (X = x)P (Y = y) = P (X = x) = pX (x) = P (Y = y) Example 32.2 Given the following table. X\ Y X=1 X=2 X=3 X=4 pY (y) Y=1 Y=2 Y=3 .01 .20 .09 .07 .00 .03 .09 .05 .06 .03 .25 .12 .2 .5 .3 pX (x) .3 .1 .2 .4 1 327 Find pX Y (x y) where Y = 2. Solution. pXY (1, 2) pY (2) pXY (2, 2) pX Y (2 2) = pY (2) pXY (3, 2) pX Y (3 2) = pY (2) pXY (4, 2) pX Y (4 2) = pY (2) pXY (x, 2) pX Y (x 2) = pY (2) pX Y (1 2) = .2 = 0.4 .5 0 =0 .5 .05 = 0.1 .5 .25 = 0.5 .5 0 = 0, x > 4 .5 = = = = = Example 32.3 If X and Y are independent Poisson random variables with respective parameters 1 and 2 , ca cu ate the conditiona distribution of X, given that X + Y = n.
328 So ution. We have P (X = k X + Y = n) = JOINT DISTRIBUTIONS
P (X = k, X + Y = n) P (X + Y = n) P (X = k, Y = n k) = P (X + Y = n) P (X = k)P (Y = n k)) = P (X + Y = n) e1 k e2 nk e(1 +2 ) (1 + 2 )n 2 1 = k! (n k)! n k)! (1 + 2 )n = n k 1 1 + 2 k 1 2 1 + 2 nk In other words, the conditiona mass distribution function of X given that X + Y = n, is the binomia distribution with parameters n and 11 2 +
32 CONDITIONAL DISTRIBUTIONS: DISCRETE CASE 329 Practice Prob ems Prob em 32.1 Given the fo owing tab e. X\ Y X=0 X=1 pY (y) Find pX Y (x y) wher e Y = 1. Problem 32.2 Choose a number X from the set {1, 2, 3, 4, 5} and then ch oose a number Y from {1, 2, , X}. (a) Find the joint mass function of X and Y. ( b) Find the conditional mass function of X given that Y = i. (c) Are X and Y ind ependent? Problem 32.3 Consider the following hypothetical joint distribution of X, a persons grade on the AP calculus exam (a number between 1 and 5), and thei r grade Y in their high school calculus course, which we assume was A = 4, B = 3 , or C = 2. X\Y 5 4 3 2 1 pY (y) 4 0.1 0.15 0.10 0 0 0.35 3 0.05 0.15 0.15 0.05 0 0.40 2 0 0 0.10 0.10 0.05 0.25 pX (x) 0.15 0.3 0.35 0.15 0.05 1 Y=0 Y=1 pX (x) .4 .1 .5 .2 .3 .5 .6 .4 1 Find P (X Y = 4) for X = 3, 4, 5. Problem 32.4 A fair coin is tossed 4 times. Le t the random variable X denote the number of heads in the rst 3 tosses, and let t he random variable Y denote the number of heads in the last 3 tosses. The joint pmf is given by the following table
330 X\Y 0 1 2 3 pY (y) JOINT DISTRIBUTIONS 0 1 2 3 1/16 1/16 0 0 1/16 3/16 2/16 0 0 2/16 3/16 1/16 0 0 1/16 1/16 2/16 6/16 6/16 2/16 pX (x) 2/16 6/16 6/16 2/16 1 What is the conditional pmf of the number of heads in the rst 3 coin tosses given exactly 1 head was observed in the last 3 tosses? Problem 32.5 Two dice are rol led. Let X and Y denote, respectively, the largest and smallest values obtained. Compute the conditional mass function of Y given X = x, for x = 1, 2, , 6. Are X and Y independent? Problem 32.6 Let X and Y be discrete random variables with joint probability function pXY (x, y) = n!y x (pe1 )y (1p)ny y!(ny)!x! 0 y = 0, 1, , n ; x = 0, 1, otherwise (a) Find pY (y). (b) Find the conditional probability distribution of X, given Y = y. Are X and Y independent? Justify your answer. Problem 32.7 Let X and Y hav e the joint probability function pXY (x, y) described as follows: X\ Y 0 1 2 pY (y) Find pX Y (x y) and pY X (y x). Problem 32.8 Let X and Y be random variable s with joint probability mass function pXY (x, y) = c(1 2x )y 0 1/18 4/18 6/18 11 /18 1 3/18 3/18 1/18 7/18 pX (x) 4/18 7/18 7/18 1
32 CONDITIONAL DISTRIBUTIONS: DISCRETE CASE 331 where x = 0, 1, , N 1 and y = 0, 1, 2, (a) Find c. (b) Find pX (x). (c) Find X (y x), the conditional probability mass function of Y given X = x. Problem 32 .9 Let X and Y be identically independent Poisson random variables with paramter . Find P (X = k X + Y = n). Problem 32.10 If two cards are randomly drawn (witho ut replacement) from an ordinary deck of 52 playing cards, Y is the number of ac es obtained in the rst draw and X is the total number of aces obtained in both dr aws, nd (a) the joint probability distribution of X and Y ; (b) the marginal dist ribution of Y ; (c) the conditional distribution of X given Y = 1.
332 JOINT DISTRIBUTIONS 33 Conditional Distributions: Continuous Case In this section, we develop the distribution of X given Y when both are continuo us random variables. Unlike the discrete case, we cannot use simple conditional probability to de ne the conditional probability of an event given Y = y, because the conditioning event has probability 0 for any y. However, we motivate our app roach by the following argument. Suppose X and Y are two continuous random varia bles with joint density fXY (x, y). Let fX Y (x y) denote the probability densit y function of X given that Y = y. We de ne b P (a < X < b Y = y) = a fX Y (x y)dx.
Then for very small we have (See Remark 22.1) P (x X x + Y = y) fX Y (x y). On t e other hand, for small we have
P (x X x + Y = y) P (x X x + y Y y + ) P (x X x + , y Y y + ) = P x, y) fY (y) In the limit, as ten s to 0, we are left with fX Y (x y) fXY (x, y) . fY (y) This suggests the following e nition. The con itional ensity function of X given Y = y is fXY (x, y) fX Y (x y) = fY (y) provided that fY (y) > 0. Compare this de nition with the discrete case where pX Y (x y) = pXY (x, y) . pY (y)
33 CONDITIONAL DISTRIBUTIONS: CONTINUOUS CASE Example 33.1 Suppose X and Y have the following joint density fXY (x, y) = X + Y < 1 0 otherwise 1 2 333
(a) Find the marginal distribution of X. 1 (b) Find the conditional distribution of Y given X = 2 . Solution. (a) Clearly, X only takes values in (1, 1). So fX ( x) = 0 if x 1. Let 1 < x < 1, 1 x 1 1 fX (x) = dy = dy = 1 x . 2 1+ x 2 (b conditional density of Y given X = fY X (y x) = 1 f ( 2 , y) = fX ( 1 ) 2 1 2 is then given by 1 1 < y < 1 2 2 0 otherwise
Thus, fY X follows a uniform distribution on the interval 1 , 1 2 2 Example 33. 2 Suppose that X is uniformly distributed on the interval [0, 1] and that, given X = x, Y is uniformly distributed on the interval [1 x, 1]. (a) Determine the j oint density fXY (x, y). 1 (b) Find the probability P (Y 2 ). Solution. Since X is uniformly istribute on [0, 1], we have fX (x) = 1, 0 x 1. Similarly, since, given X = x, Y is uniformly istribute on [1 x, 1], the 1 1 con itional ensit y of Y given X = x is 1(1x) = x on the interval [1 x, 1]; 1 i.e., fY X (y x) = x , 1 x y 1 for 0 x 1. Thus fXY (x, y) = fX (x)fY X (y x) = 1 , 0 < x < 1, 1 x < y < 1 x
334 (b) Using Figure 33.1 we 0 1 2 JOINT DISTRIBUTIONS 1 1x 0 1 2 1 y x + x 1 1 2 1 2 1 1 y x x 1 (1 x) x + x 1 1 2 1/2 x x = 1 + ln 2 2 Figure 33.1 Note that fX Y (x y)dx =
nd 1 P (Y = 2 =
fXY (x, y) fY (y) dx = = 1. fY (y) fY (y) The conditional cumulative distribution function of X given Y = y is de ned by x FX Y (x y) = P (X x Y = y) = fX Y (t y)dt. From this de nition, it follows fX Y (x y) = FX Y (x y). x Example 33.3 The joint density of X and Y is given by fXY (x, y) = 15 x(2 2 x y) 0 x, y 1 0 otherwise Compute the conditional density of X, given that Y = y for 0 y 1.

33 CONDITIONAL DISTRIBUTIONS: CONTINUOUS CASE Solution. The marginal density fun ction of Y is 1 335 fY (y) = 0 15 15 x(2 x y)dx = 2 2 2 y 3 2 . Thus, fX Y (x y) = fXY (x, y) fY (y) x(2 x y) = 2 y 3 2 6x(2 x y) = 4 3y Example 33.4 The joint density function of X and Y is given by e x y ye fXY (x, y) = Compute P (X > 1 Y = y). y 0 x 0, y 0 otherwise Solution. The marginal fY (y) = e Thus, y 0 x 1 x e y x = ey e y y 0 = ey . fX Y (x y) = = fXY (x, y) fY (y) e x y ye y ey 1 x = e y y
ensity function of Y is
336 Hence, JOINT DISTRIBUTIONS P (X > 1 Y = y) = 1 1 x e y dx y x 1 = e y = e y 1 We end this section with the following theorem. Theorem 33.1 Continuous random variables X and Y are independent if and only if fX Y (x y) = fX (x). Proof. Suppose rst that X and Y are independent. Then fXY (x , y) = fX (x)fY (y). Thus, fX (x)fY (y) fXY (x, y) = = fX (x). fX Y (x y) = fY ( y) fY (y) Conversely, suppose that fX Y (x y) = fX (x). Then fXY (x, y) = fX Y ( x y)fY (y) = fX (x)fY (y). This shows that X and Y are independent Example 33.5 Let X and Y be two continuous random variables with joint density function fXY ( x, y) = c 0y<x2 0 otherwise (a) Find fX (x), fY (y) and fX Y (x 1). (b) Are X and Y independent? Solution. ( a) We have fX (x) = 0 2 x cdy = cx, 0 x 2 cdx = c(2 y), 0 y 2 fY (y) = y and fX Y (x 1) = fXY (x, 1) c = = 1, 0 x 1. fY (1) c (b) Since fX Y (x 1) = fX (x), X and Y are dependent
33 CONDITIONAL DISTRIBUTIONS: CONTINUOUS CASE 337 Practice Problems Problem 33.1 Let X and Y be two random variables with joint density function fXY (x, y) = 5x2 y 1 x 1, y x 0 otherwise Find fX Y (x y), the conditional probability density function of X given Y = y. Sketch the graph of fX Y (x 0.5). Problem 33.2 Suppose that X and Y have joint d istribution fXY (x, y) = 8xy 0 x < y 1 0 otherwise Find fX Y (x y), the conditional probability density function of X given Y = y. Problem 33.3 Suppose that X and Y have joint distribution fXY (x, y) = 3y 2 x3 0 0y<x1 otherwise Find fX Y (x y), the conditional probability density function of Y given X = x. Problem 33.4 The joint density function of X and Y is given by fXY (x, y) = xex(y +1) x 0, y 0 0 otherwise Fin the con itional ensity of X given Y = y an that of Y given X = x. Problem 33.5 Let X an Y be continuous ran om variables with con itional an marginal p . .f.s given by x3 ex I(0,) (x) fX (x) = 6
338 an fX Y (x y) = JOINT DISTRIBUTIONS 3y 2 I(0,x) . x3 (a) Find the joint p.d.f. of X and Y. (b) Find the conditional p.d.f. of X given Y = y. Problem 33.6 Suppose X, Y are two continuous random variables with joint probability density function fXY (x, y) = 12xy(1 x) 0 < x, y < 1 0 otherwise (a) Find fX Y (x y). Are X and Y independent? (b) Find P (Y < 1 X > 1 ). 2 2 Pr oblem 33.7 The joint probability density function of the random variables X and Y is given by 1 x y + 1 1 x 2, 0 y 1 3 fXY (x, y) = 0 otherwise (a) Find the con ditional probability density function of X given Y = y. 1 (b) Find P (X < 3 Y = 2 ). 2 Problem 33.8 Let X and Y be continuous random variables with joint densi ty function fXY (x, y) = Calculate P Y < X X = 1 3 24xy 0 < x < 1, 0 < y < 1 x 0 otherwise . Problem 33.9 Once a re is reported to a re insurance company, the company makes an initial estimate, X, of the amount it will pay to the claimant for the re loss. When the claim is nally settled, the company pays an amount, Y, to the claimant. The company has determined that X and Y have the joint density function 2 y (2x1)/ (x1) x > 1, y > 1 x2 (x1) fXY (x, y) = 0 otherwise Given that the initial claim es timated by the company is 2, determine the probability that the nal settlement am ount is between 1 and 3 .
33 CONDITIONAL DISTRIBUTIONS: CONTINUOUS CASE 339 Problem 33.10 A company o ers a basic life insurance policy to its employees, as w ell as a supplemental life insurance policy. To purchase the supplemental policy , an employee must rst purchase the basic policy. Let X denote the proportion of employees who purchase the basic policy, and Y the proportion of employees who p urchase the supplemental policy. Let X and Y have the joint density function fXY (x, y) = 2(x + y) on the region where the density is positive. Given that 10% o f the employees buy the basic policy, what is the probability that fewer than 5% buy the supplemental policy? Problem 33.11 An auto insurance policy will pay fo r damage to both the policyholders car and the other drivers car in the event that the policyholder is responsible for an accident. The size of the payment for da mage to the policyholders car, X, has a marginal density function of 1 for 0 < x < 1. Given X = x, the size of the payment for damage to the other drivers car, Y, has conditional density of 1 for x < y < x + 1. If the policyholder is responsi ble for an accident, what is the probability that the payment for damage to the other drivers car will be greater than 0.5? Problem 33.12 You are given the follo wing information about N, the annual number of claims for a randomly selected in sured: 1 2 1 P (N = 1) = 3 1 P (N > 1) = 6 P (N = 0) = Let S denote the total an nual claim amount for an insured. When N = 1, S is exponentially distributed wit h mean 5 . When N > 1, S is exponentially distributed with mean 8 . Determine P (4 < S < 8). Problem 33.13 Let Y have a uniform distribution on the interval (0, 1), and let the con ditional distribution of X given Y = y be uniform on the int erval (0, y). What is the marginal density function of X for 0 < x < 1?
340 JOINT DISTRIBUTIONS Problem 33.14 Suppose that X has a continuous distribution with p.d.f. fX (x) = 2x on (0, 1) and 0 elsewhere. Suppose that Y is a continuous random variable suc h that the conditional distribution of Y given X = x is uniform on the interval (0, x). Find the mean and variance of Y. Problem 33.15 The distribution of Y, gi ven X, is uniform on the interval [0, X]. The marginal density of X is 2x for 0 < x < 1 fX (x) = 0 otherwise. Determine the conditional density of X, given Y = y > 0.
34 JOINT PROBABILITY DISTRIBUTIONS OF FUNCTIONS OF RANDOM VARIABLES341 34 Joint Probability Distributions Theorem 28.1 provided a result for able: if Y = g(X) is a function of and di erentiable then the pdf of y of Functions of Random Variables nding the pdf of a function of one random vari the random variable X, where g(x) is monotone Y is given by fY (y) = fX (g 1 (y)) d 1 g (y) . d
An extension to functions of two random variables is given in the following theo rem. Theorem 34.1 Let X and Y be jointly continuous random variables with joint probability density function fXY (x, y). Let U = g1 (X, Y ) and V = g2 (X, Y ). Assume that the functions u = g1 (x, y) and v = g2 (x, y) can be solved uniquely for x and y. Furthermore, suppose that g1 and g2 have continuous partial deriva tives at all points (x, y) and such that the Jacobian determinant g1 x g2 x g1 y g2 y J(x, y) = = g1 g2 g1 g2 =0 x y y x for all x and y. Then the random variables U and V are continuous random variabl es with joint density function given by fU V (u, v) = fXY (x(u, v), y(u, v)) J(x (u, v), y(u, v)) 1 Proof. We rst recall the reader about the change of variable fo rmula for a double integral. Suppose x = x(u, v) and y = y(u, v) are two di erenti able functions of u and v. We assume that the functions x and y take a point in the uvplane to exactly one point in the xyplane. Let us see what happens to a smal l rectangle T in the uvplane with sides of lengths u and v as shown in Figure 34.1.
Figure 34.1 Since the side lengths are small, by local linearity each side of th e rectangle in the uvplane is transformed into a line segment in the xyplane. The result is that the rectangle in the uvplane is transformed into a parallelogram R in the xyplane with sides in vector form are a = [x(u + u, v) x(u, v)]i + [y(u + u , v) y(u, v)]j and b = [x(u, v + v) x(u, v)]i + [y(u, v + v) y(u, v)]j Now, the ea of R is Area R a b = x y x y uv u v v u (x,y) , (u,v) y x ui + uj u u x y v i + v j v v Using determinant notation, we de ne the Jacobian, (x, y) x y x y = = (u, v) u v e can write Area R (x, y) uv (u, v) x u y u as follows x v y v
34 JOINT PROBABILITY DISTRIBUTIONS OF FUNCTIONS OF RANDOM VARIABLES343 Now, supp ose we are integrating f (x, y) over a region R. Partition R into mn small paral lelograms. Then using Riemann sums we can write m n f (x, y)dxdy R j=1 i=1 m n f (xij , yij ) Area of Rij f (uij , vij ) j=1 i=1 (x, y) uv (u, v) where (xij , yij ) in Rij corresponds to a point (uij , vij ) in Tij . Now, lett ing m, n to otbain f (x, y)dxdy = R T f (x(u, v), y(u, v)) (x, y) dudv. (u, v) The result of the theorem follows from the fact that if a region R in the xyplane maps into the region T in the uvplane then we must have P ((X, Y ) R) = R fXY (x, y)dxdy fXY (x(u, v), y(u, v)) J(x(u, v), y(u, v)) 1 dudv T = =P ((U, V ) T ) Example 34.1 Let X and Y be jointly continuous random variables with density function fXY (x, y). Let U = X + Y and V = X Y. Find the joint dens ity function of U and V. Solution. Let u = g1 (x, y) = x+y and v = g2 (x, y) = xy . Then x = Moreover 1 1 J(x, y) = = 2 1 1 Thus, 1 fU V (u, v) = fXY 2 u+v uv , 2 2 u+v 2 uv . 2 and y =
344 JOINT DISTRIBUTIONS Example 34.2 Let X and Y be jointly continuous random variables with density fun ction x2 +y 2 1 fXY (x, y) = 2 e 2 . Let U = X + Y and V = X Y. ind the joint den sity function of U and V. Solution. Since J(x, y) = 2 we have fU V (u, v) = 1 ( u +v )2 +( uv )2 1 u2 +v2 2 2 2 e e 4 = 4 4 Exam le 34.3 Su ose that X and Y have joint density function given by fXY (x, y ) = 4xy 0 < x < 1, 0 < y < 1 0 otherwise Let U = X and V = XY. Y (a) ind the joint density function of U and V. (b) ind the marginal density of U and V . (c) Are U and V inde endent? Solution. (a) No w, if u = g1 (x, y) = we nd x = uv and y = x y v . u and v = g2 (x, y) = xy then solving for x and y Moreover, yx2 y x 1 y J(x, y) = By Theorem 34.1, we nd fU V (u, v) = 1 fXY ( uv, 2u = 2x y v 2v v ) = , 0 < uv < 1, 0 < < 1 u u u and 0 otherwise. The region where fU V is shown in density of U is u fU (u) = 0 1 u 2v dv = u, u 1 u 2v 1 dv = 3 , u > 1 u u fU (u) = 0
igure 34.2. (b) The marginal
34 JOINT PROBABILITY DISTRIBUTIONS O arginal density of V is 1 v fV (v) = 0 fU V (u, v)du = v 2v du = 4v ln v, 0 < v < 1 u
UNCTIONS O RANDOM VARIABLES345 and the m
(c) Since fU V (u, v) = fU (u)fV (v), U and V are de endent
igure 34.2
346 JOINT DISTRIBUTIONS Practice Problems Problem 34.1 Let X and Y be two random variables with joint df fXY . Let Z = aX + bY and W = cX + dY where ad bc = 0. ind the joint robability density functi on of Z and W. Problem 34.2 Let X1 and X2 be two inde endent ex onential random variables each having arameter . Find the joint density function of Y1 = X1 + X2 and Y2 = eX2 . Prob em 34.3 Let X and Y be random variab es with joint pdf fXY (x, y). Let R = X 2 + Y 2 Y and = tan1 X with < . ind fR (r, ). Problem 34.4 nd be two random variables with joint pd XY (x, y). Let Z = Y Y g(X, Y ) = X 2 + Y 2 and W = X . ind ZW (z, w). Problem 34.5 I X and Y are independent gamma random variables with parameters (, ) and (, ) respective y, compute the joint dens ity of U = X + Y and V = X . X+Y Prob em 34.6 Let X1 and X2 be two continuous ra ndom variab es with joint density function e(x1 +x2 ) x1 0, x2 0 fX1 X2 (x1 , x2 ) = 0 otherwise Let Y1 = X1 + X2 and Y2 = Y2 . X1 . X1 +X2 Find the joint density function of Y1 and Prob em 34.7 Let X1 and X2 be two independent norma random variab es with param eters (0,1) and (0,4) respective y. Let Y1 = 2X1 + X2 and Y2 = X1 3X2 . Find fY1 Y2 (y1 , y2 ). Prob em 34.8 Let X be a uniform random variab e on (0, 2) and Y a n ex onential random variable with = 1 and independent of X. Show that

34 JOINT PROBABILITY DISTRIBUTIONS OF FUNCTIONS OF RANDOM VARIABLES347 U= 2Y cos X and V = 2Y sin X are independent standard norma random variab es Prob em 34.9 Let X and Y be two random variab es with joint density function fXY . Compute the pdf of U = X + Y . What is the pdf in the case X and Y are independent? Hint: et V = Y. Prob em 34.10 Let X and Y be two random variab es with joint density function fXY . Comp ute the pdf of U = Y X. Prob em 34.11 Let X and Y be two random variab es with j oint density function fXY . Compute the pdf of U = XY. Hint: et V = X. Prob em 34.12 The dai y amounts of Coke and Diet Coke so d by a vendor are independent a nd fo ow exponentia distributions with a mean of 1 (units are 100s of ga ons) . Use Theorem 34.1 to obtain the distribution of the ratio of Coke to Diet Coke so d.
Properties of Expectation We have seen that the expected va ue of a random variab e is a weighted average of the possib e va ues of X and a so is the center of the distribution of the va riab e. Reca that the expected va ue of a discrete random variab e X with prob abi ity mass function p(x) is de ned by E(X) = x xp(x) provided that the sum is nite. For a continuous random variab e X with probabi it y density function f (x), the expected va ue is given by E(X) = xf (x)dx provided that the improper integra is convergent. In this chapter we deve op an d exp oit properties of expected va ues. 35 Expected Va ue of a Function of Two Random Variab es In this section, we earn some equa ities and inequa ities about the expectation of random variab es. Our goa s are to become comfortab e with the expectation o perator and earn about some usefu properties. First, we introduce the de nition of expectation of a function of two random variab es: Suppose that X and Y are t wo random variab es taking va ues in SX and SY respective y. For a function g : SX SY R the expected va ue of g(X, Y ) is E(g(X, Y ) = g(x, y)pXY (x, y). xSX ySY 349
350 PROPERTIES OF EXPECTATION if X and Y are discrete with joint probabi ity mass function pXY (x, y) and E(g(X, Y )) = g(x, y)fXY (x, y)dxdy if X and Y are continuous with joint probabi ity density function fXY (x, y). Ex amp e 35.1 Let X and Y be two discrete random variab es with joint probabi ity m ass function: pXY (1, 1) = 1 , pXY (1, 2) = 1 , pXY (2, 1) = 1 , pXY (2, 2) = 3 8 2 Find the expected va ue of g(X, Y ) = XY. So ution. The expected va ue of th e function g(X, Y ) = XY is ca cu ated as fo ows: 2 2 1 24 E(g(X, Y )) =E(XY ) = x=1 y=1 xypXY (x, y) 1 1 1 1 =(1)(1)( ) + (1)(2)( ) + (2)(1)( ) + (2)(2)( ) 3 8 2 24 7 = 4 An importa nt app ication of the above de nition is the fo owing resu t. Proposition 35.1 Th e expected va ue of the sum/di erence of two random variab es is equa to the sum/ di erence of their expectations. That is, E(X + Y ) = E(X) + E(Y ) and E(X Y ) = E (X) E(Y ).
35 EXPECTED VALUE OF A FUNCTION OF TWO RANDOM VARIABLES351 Proof. We prove the r esu t for discrete random variab es X and Y with joint probabi ity mass function pXY (x, y). Letting g(X, Y ) = X Y we have E(X Y ) = x y (x y)pXY (x, y) xpXY (x, y) x y x y = = x ypXY (x, y) y y x x y pXY (x, y) ypY (y) y pXY (x, y) = x xpX (x)
=E(X) E(Y ) A simi ar proof ho ds for the continuous case where you just need to rep ace the sums by improper integra s and the joint probabi ity mass function by the joint probabi ity density function Using mathematica induction one can e asi y extend the previous resu t to E(X1 + X2 + + Xn ) = E(X1 ) + E(X2 ) + + E( n ), E(Xi ) < . Example 35.2 During commencement at the United States Naval acade my, there is a tradition that everyone throws their hat into the air at the conc lusion of the ceremony. Later that day, the hats are collected and each graduate is given a hat at random. Assuming that all of the hats are indistinguishable f rom the others( so hats are really given back at random) and that there are 1000 graduates, calculate E(X), where X is the number of people who receive the hat that they wore during the ceremony. Solution. Let Xi = 0 if the ith person does not get his/her hat 1 if the ith person gets his/her hat Then X = X1 + X2 + X1000 and E(X) = E(X1 ) + E(X2 ) + + E(X1000 ).
352 But for 1 i 1000 E(Xi ) = 0 Hence, PROPERTIES OF EXPECTATION 999 1 1 +1 = . 1000 1000 1000 E(X) = 1000E(Xi ) = 1 Example 35.3 (Sample Mean) Let X1 , X2 , , Xn be a sequenc e of independent and identically distributed random variables, each having a mea n and variance 2 . De ne a new random variable by X1 + X2 + + X n X= . n We call the sample mean. Find E(X). Solution. The expected value of X is E(X) = E X1 + X 2 + + X n 1 = n n n E(Xi ) = . i=1 Because of this result, when the distribution mean is unknown, the sample mean i s often used in statisitcs to estimate it The following property is known as the monotonicity property of the expected value. Proposition 35.2 If X is a nonnega tive random variable then E(X) 0. Thus, if X an Y are two ran om variables such that X Y then E(X) E(Y ). Proof. We prove the result for the continuous case. W e have E(X) = xf (x) x xf (x) x 0 0 =
35 EXPECTED VALUE OF A FUNCTION OF TWO RANDOM VARIABLES353 since f (x) 0 so the integran is nonnegative. Now, if X Y then X Y 0 so that by the previous proposi tion we can write E(X) E(Y ) = E(X Y ) 0 As a irect application of the monotoni city property we have Proposition 35.3 (Booles Inequality) For any events A1 , A2 , , An we have n n P i=1 Ai i=1 P (Ai ). Proof. For i = 1, , n de ne Xi = Let X= i=1 1 if Ai occurs 0 otherwise n Xi so X denotes the number of the events Ai that occur. Also, let Y = 1 if X 1 occu rs 0 otherwise so Y is equal to 1 if at least one of the Ai occurs an Y so that E(X) E(Y ). But n n E(X) = i=1 E(Xi ) = i=1 P (Ai ) an E(Y ) = P { at least one of the Ai occur } = P ( Thus, the result follows. N ote that for any set A we have E(IA ) = IA (x)f (x) x = A n i=1 Ai ) . f (x) x = P (A)
0 otherwise. Clearly, X
354 PROPERTIES OF EXPECTATION
Proposition 35.4 If X is a ran om variable with range [a, b] then a E(X) b. Proo f. Let Y = X a 0. Then E(Y ) 0. But E(Y ) = E(X)E(a) = E(X)a 0. Thus, E(X) a. Si arly, let Z = bX 0. Then E(Z) = bE(X) 0 or E(X) b We have etermine that the expe ctation of a sum is the sum of the expectations. The same is not always true for pro ucts: in general, the expectation of a pro uct nee not equal the pro uct o f the expectations. But it is true in an important special case, namely, when th e ran om variables are in epen ent. Proposition 35.5 If X an Y are in epen ent ran om variables then for any function h an g we have E(g(X)h(Y )) = E(g(X))E(h (Y )) In particular, E(XY ) = E(X)E(Y ). Proof. We prove the result for the cont inuous case. The proof of the iscrete case is similar. Let X an Y be two in ep en ent ran om variables with joint ensity function fXY (x, y). Then E(g(X)h(Y )) = g(x)h(y)fXY (x, y) x y g(x)h(y)fX (x)fY (y) x y = = h(y)fY (y) y g(x)fX (x) x =E(h(Y ))E(g(X)) We next give a simple example to show that the expecte values nee not multiply if the ran om variables are not in epen ent. Example 35.4 Cons i er a single toss of a coin. We e ne the ran om variable X to be 1 if hea s turn s up an 0 if tails turns up, an we set Y = 1 X. Thus X an Y are epen ent. Sh ow that E(XY ) = E(X)E(Y ).

35 EXPECTED VALUE OF A FUNCTION OF TWO RANDOM VARIABLES355 Solution. Clearly, E( X) = E(Y ) = 1 . But XY = 0 so that E(XY ) = 0 = E(X)E(Y ) 2 Example 35.5 Suppos e a box contains 10 green, 10 re an 10 black balls. We raw 10 balls from the box by sampling with replacement. Let X be the number of green balls, an Y be t he number of black balls in the sample. (a) Fin E(XY ). (b) Are X an Y in epen ent? Explain. Solution. First we note that X an Y are binomial with n = 10 an p = 1 . 3 (a) Let Xi be 1 if we get a green ball on the ith raw an 0 otherwis e, an Yj be the event that in j th raw we got a black ball. Trivially, Xi an Yj are in epen ent if 1 i = j 10. Moreover, Xi Yi = 0 for all 1 i 10. Since X = X1 + X2 + X10 and Y = Y1 + Y2 + Y10 we have XY = 1i=j10 Xi Yj . Hence, E(XY ) = 1i=j10 E(Xi Yj ) = 10 , 3 1 1 E(Xi )E(Yj ) = 90 = 10. 3 3 1i=j10 (b) Since E(X) = E(Y ) = are dependent we have E(XY ) = E(X)E(Y ) so X and Y The following inequality will be of importance in the next section Proposition 3 5.6 (Markovs Inequality) If X 0 an c > 0 then P (X c) E(X) . c Proof. Let c > 0. De ne I= 1 if X c 0 otherwise Since X 0, we have I X . Taking expectations of both si e we the result follows since E(I) = P (X c) c n c E(I) E(X) . Now

356 PROPERTIES OF EXPECTATION Example 35.6 Let X be a non negative ran om variable. Let a be a positive consta nt. tX Prove that P (X a) E(eta ) for all t 0. e Solution. Applying Markovs inequ ality we n P (X a) = P (tX ta) = P (etX eta ) E(etX ) eta As an important application of the previous result we have Proposition 35.7 If X 0 an E(X) = 0 then P (X = 0) = 1. Proof. Since E(X) = 0, by the previous resul t we n P (X c) = 0 for all c > 0. But 1 1 P (X > 0) = P (X > ) P (X > ) = 0. n n n=1 n=1 Hence, P (X > 0) = 0. Since X 0, we have 1 = P (X 0) = P (X = 0) + P (X > 0) = P (X = 0) Corollary 35.1 Let X be a ran om variable. If V ar(X) = 0, the n P (X = E(X)) = 1. Proof. Suppose that V ar(X) = 0. Since (X E(X))2 0 an V ar( X) = E((X E(X))2 ), by the previous result we have P (X E(X) = 0) = 1. That is, P (X = E(X)) = 1 Example 35.7 (expecte value of a Binomial Ran om Variable) Let X be a a binomial ran om variable with parameters (n, p). Fin E(X). Solution. We have that X is the number of successes in n trials. For 1 i n let Xi enote t he number of successes in the ith trial. Then E(Xi ) = 0(1p)+1p = p. Since X = X1 + X2 + + Xn , we nd E(X) = n E(Xi ) = n p = i=1 i=1 np

35 EXPECTED VALUE OF A FUNCTION OF TWO RANDOM VARIABLES357 Practice Problems Problem 35.1 Let X and Y be independent random variables, both being equally lik ely to be any of the numbers 1, 2, , m. Find E( X Y ). Problem 35.2 Let X and Y be random variables with joint pdf fXY (x, y) = Find E(XY ). Problem 35.3 Let X and Y be two independent uniformly distributed random variables in [0,1]. Find E( X Y ). Problem 35.4 Let X and Y be continuous random variables with joint pd f fXY (x, y) = Find E(X 2 Y ) and E(X 2 + Y 2 ). Problem 35.5 Suppose that E(X) = 5 and E(Y ) = 2. Find E(3X + 4Y 7). Problem 35.6 Suppose that X and Y are indep endent, and that E(X) = 5, E(Y ) = 2. Find E[(3X 4)(2Y + 7)]. Problem 35.7 An acc ident occurs at a point X that is uniformly distributed on a road of length L. A t the time of the accident an ambulance is at location Y that is also uniformly distributed on the road. Assuming that X and Y are independent, nd the expected d istance between the ambulance and the point of the accident. 2(x + y) 0 < x < y < 1 0 otherwise 1 0 < x < 1, x < y < x + 1 0 otherwise
358 PROPERTIES OF EXPECTATION Problem 35.8 A group of N people throw their hats into the center of a room. The hats are mixed, and each person randomly selects one. Find the expected number of people that select their own hat. Problem 35.9 Twenty people, consisting of 1 0 married couples, are to be seated at ve di erent tables, with four people at each table. If the seating is done at random, what is the expected number of married couples that are seated at the same table? Problem 35.10 Suppose that A and B e ach randomly, and independently, choose 3 out of 10 objects. Find the expected n umber of objects (a) chosen by both A and B. (b) not chosen by either A or B (c) chosen exactly by one of A and B. Problem 35.11 If E(X) = 1 and Var(X) = 5 nd (a ) E[(2 + X)2 ] (b) Var(4 + 3X) Problem 35.12 Let T1 be the time between a car ac cident and reporting a claim to the insurance company. Let T2 be the time betwee n the report of the claim and payment of the claim. The joint density function o f T1 and T2 , f (t1 , t2 ), is constant over the region 0 < t1 < 6, 0 < t2 < 6, t1 + t2 < 10, and zero otherwise. Determine E[T1 +T2 ], the expected time betwee n a car accident and payment of the claim. Problem 35.13 Let T1 and T2 represent the lifetimes in hours of two linked components in an electronic device. The jo int density function for T1 and T2 is uniform over the region de ned by 0 t1 t2 L, where L is a positive constant. Determine the expected value of the sum of the squares of T1 and T2 . Problem 35.14 Let X and Y be two independent random varia bles with X = 1, Y = 1 2 2 1, X = 2 , and Y = 2. Compute E[(X + 1)2 (Y 1)2 ].
36 COVARIANCE, VARIANCE OF SUMS, AND CORRELATIONS 359 36 Covariance, Variance of Sum , and Correlation So far, We have di cu ed the ab ence or pre ence of a relation hip between two random variable , i.e. independence or dependence. But if there i in fact a rel ation hip, the relation hip may be either weak or trong. For example, if X i t he weight of a ample of water and Y i the volume of the ample of water then t here i a trong relation hip between X and Y. On the other hand, if X i the we ight of a per on and Y denote the ame per on height then there i a relation h ip between X and Y but not a trong a in the previou example. We would like a mea ure that can quantify thi di erence in the trength of a relation hip betwee n two random variable . The covariance between X and Y i de ned by Cov(X, Y ) = E [(X E(X))(Y E(Y ))]. An alternative expre ion that i ometime more convenient i Cov(X, Y ) =E(XY E(X)Y XE(Y ) + E(X)E(Y )) =E(XY ) E(X)E(Y ) E(X)E(Y ) + E(X )E(Y ) =E(XY ) E(X)E(Y ) Recall that for independent X, Y we have E(XY ) = E(X)E (Y ) and o Cov(X, Y ) = 0. However, the conver e tatement i fal e a there ex i t random variable that have covariance 0 but are dependent. For example, let X be a random variable uch that P (X = 0) = P (X = 1) = P (X = 1) = and de ne Y = 0 if X = 0 1 otherwi e 1 3 Thu , Y depend on X. Clearly, XY = 0 o that E(XY ) = 0. Al o, 1 E(X) = (0 + 1 1) = 0 3
PROPERTIES OF EXPECTATION Cov(X, Y ) = E(XY ) E(X)E(Y ) = 0. U eful fact are collected in the next re ult . Theorem 36.1 (a) Cov(X, Y ) = Cov(Y, X) (Symmetry) (b) Cov(X, X) = V ar(X) (c) Cov(aX, Y ) = aCov(X, Y ) n m n m (d) Cov i=1 Xi , j=1 Yj = i=1 j=1 Cov(Xi , Yj ) Proof. (a) Cov(X, Y ) = E(XY )E(X)E(Y ) = E(Y X)E(Y )E(X) = Cov(Y, X). (b) Cov( X, X) = E(X 2 ) (E(X))2 = V ar(X). (c) Cov(aX, Y ) = E(aXY ) E(aX)E(Y ) = aE(XY ) aE(X)E(Y ) = a(E(XY ) E(X)E(Y )) = aCov(X, Y ). m m (d) Fir t note that E [ n Xi ] = n E(Xi ) and E j=1 E(Yj ). j=1 Yj = i=1 i=1 Then n m n n m m Cov i=1 Xi , j=1 Yj =E i=1 n Xi i=1 E(Xi ) j=1 m Yj j=1 E(Yj ) =E i=1 n (Xi E(Xi )) j=1 m (Yj E(Yj )) =E i=1 j=1 n m (Xi E(Xi ))(Yj E(Yj )) E[(Xi E(Xi ))(Yj E(Yj ))] = i=1 j=1 n m = i=1 j=1 Cov(Xi , Yj ) Example 36.1 Given that E(X) = 5, E(X 2 ) = 27.4, E(Y ) = 7, E(Y 2 ) = 51.4 and Var(X + Y ) = 8, nd Cov(X + Y, X + 1.2Y ).
360 and thu
36 COVARIANCE, VARIANCE OF SUMS, AND CORRELATIONS Solution. By de nition, 361 Cov(X + Y, X + 1.2Y ) = E((X + Y )(X + 1.2Y )) E(X + Y )E(X + 1.2Y ) U ing the p ropertie of expectation and the given data, we get E(X + Y )E(X + 1.2Y ) =(E(X) + E(Y ))(E(X) + 1.2E(Y )) = (5 + 7)(5 + (1.2) 7) = 160.8 E((X + Y )(X + 1.2Y )) =E(X 2 ) + 2.2E(XY ) + 1.2E(Y 2 ) =27.4 + 2.2E(XY ) + (1.2)(51.4) = 2.2E(XY ) + 89.08 Thus, Cov(X + Y, X + 1.2Y ) = 2.2E(XY ) + 89.08 160.8 = 2.2E(XY ) 71.72 T o complete the calculation, it remains to nd E(XY ). To this end we make use of t he still unused relation Var(X + Y ) = 8 8 =Var(X + Y ) = E((X + Y )2 ) (E(X + Y ))2 =E(X 2 ) + 2E(XY ) + E(Y 2 ) (E(X) + E(Y ))2 =27.4 + 2E(XY ) + 51.4 (5 + 7) 2 = 2E(XY ) 65.2 so E(XY ) = 36.6. Substituting this above gives Cov(X + Y, X + 1.2Y ) = (2.2)(36.6) 71.72 = 8.8 Example 36.2 On reviewing data on smoking (X, n umber of packages of cigarettes smoked per year) income (Y, in thousands per yea r) and health (Z number of visits to the family physician per year) for a sample of males it is found that E(X) = 10, Var(X) = 25, E(Y ) = 50, Var(Y ) = 100, E( Z) = 6, Var(Z) = 4, Cov(X, Y ) = 10, and Cov(X, Z) = 3.5. Dr. Davis, a young sta tistician, attempts to describe the variable Z in terms of X and Y by the relati on Z = X +cY , where c is a constant to be determined. Dr. Davis methodology for determining c is to nd the value of c for which Cov(X, Z) = 3.5 when Z is replace d by X + cY. What value of c does Dr. Davis nd? Solution. We have Cov(X, Z) =Cov( X, X + cY ) = Cov(X, X) + cCov(X, Y ) =Var(X) + cCov(X, Y ) = 25 + c(10) = 3.5
362 Solving for c we
nd c = 2.15
PROPERTIES OF EXPECTATION Using (b) and (d) in the previous theorem with Yj = Xj , j = 1, 2, , n we n n n V ar i=1 Xi =Cov i=1 n n Xi , i=1 Xi = i=1 i=1 n Cov(Xi , Xj ) V ar(Xi ) + i=1 i=j = Cov(Xi , Xj ) Since each pair of indices i = j appears twice in the double summation, the abov e reduces to n n V ar i=1 Xi = i=1 V ar(Xi ) + 2 i<j Cov(Xi , Xj ). In particular, if X1 , X2 , , Xn are pairwise independent then n n V ar i=1 Xi = i=1 V ar(Xi ). nd
Example 36.3 The pro t for a new product is given by Z = 3X Y 5, where X and Y are independent random variables with Var(X) = 1 and Var(Y ) = 2. What is the varia nce of Z? Solution. Using the properties of a variance, and independence, we get Var(Z) =Var(3X Y 5) = Var(3X Y ) =Var(3X) + Var(Y ) = 9Var(X) + Var(Y ) = 11 Exa mple 36.4 An insurance policy pays a total medical bene t consisting of a part pai d to the surgeon, X, and a part paid to the hospital, Y, so that the total
36 COVARIANCE, VARIANCE OF SUMS, AND CORRELATIONS 363 bene t is X + Y. Suppose that Var(X) = 5, 000, Var(Y ) = 10, 000, and Var(X + Y ) = 17, 000. If X is increased by a at amount of 100, and Y is increased by 10%, wh at is the variance of the total bene t after these increases? Solution. We need to compute Var(X + 100 + 1.1Y ). Since adding constants does not change the varian ce, this is the same as Var(X + 1.1Y ), which expands as follows: Var(X + 1.1Y ) =Var(X) + Var(1.1Y ) + 2Cov(X, 1.1Y ) =Var(X) + 1.21Var(Y ) + 2(1.1)Cov(X, Y ) We are given that Var(X) = 5, 000, Var(Y ) = 10, 000, so the only remaining unkn own quantity is Cov(X, Y ), which can be computed via the general formula for Va r(X + Y ) : 1 Cov(X, Y ) = (Var(X + Y ) Var(X) Var(Y )) 2 1 = (17, 000 5, 000 10 , 000) = 1, 000 2 Substituting this into the above formula, we get the answer: V ar(X + 1.1Y ) = 5, 000 + 1.21(10, 000) + 2(1.1)(1, 000) = 19, 300 Example 36.5 L et X be the sample mean of n independent random variables X1 , X2 , , Xn . Find V ar(X). Solution. By independence we have 1 V ar(X) = 2 n n V ar(Xi ) i=1 The following result is known as the Cauchy Schwartz inequality. Theorem 36.2 Le t X and Y be two random variables. Then Cov(X, Y )2 V ar(X)V ar(Y )
364 PROPERTIES OF EXPECTATION with equality if and only if X and Y are linearly related, i.e., Y = aX + b for some constants a and b with a = 0. Proof. Let (X, Y ) = Cov(X, Y ) V a (X)V a (Y ) .
We need to show that 1 or equivalently 1 (X, Y ) 1. If we let 2 2 X and Y den he variance of X and Y re pectively then we have 0 V ar Y X + X Y V ar(X) V ar(Y ) 2Cov(X, Y ) = + + 2 2 X Y X Y =2[1 + (X, Y )] implying that 1 (X, Y ). Simila ly, 0 V a Y X X Y V ar(X) V ar(Y ) 2Cov(X, Y ) = 2 X Y X Y =2[1 (X, Y )]
or 0. Thi implie Conver ely, uppo e that Y = a + bX. Then (X, Y ) = E(aX + bX 2 ) E(X)E(a + bX) V a (X)b2 V a (X) = bV a (X) = sign(b). b V ar(X)
implying that (X, lies that eithe X implie that Y = Y = C for ome con 1 then V a X +
Y ) 1. Suppose now that Cov(X, Y )2 = V a (X)V a (Y ). This imp Y (X, Y ) = 1 o (X, Y ) = 1. If (X, Y ) = 1 then V a X Y = 0. a + bX tant C (See Corollary 35.4) Y Y X Y where b = X > 0. If (X, Y ) = Y = Y X Y that X + Y = C or Y = a + bX where b = X < 0. X X
36 COVARIANCE, VARIANCE OF SUMS, AND CORRELATIONS If b > 0 then (X, Y ) = 1 and i f b < 0 then (X, Y ) = 1 365 The Co elation coe cient of two andom va iables X and Y is de ned by Cov(X, Y ) . ( X, Y ) = V a (X)V a (Y ) F om the above theo em we have the co elation inequali ty 1 1. The co elation coe cient is a measu e of the deg ee of linea ity between X and Y . A value of (X, Y ) nea +1 o 1 indicates a high deg ee of linea ity betw een X and Y, whe eas a value nea 0 indicates a lack of such linea ity. Co elat ion is a scaled ve sion of cova iance; note that the two pa amete s always have the same sign (positive, negative, o 0). When the sign is positive, the va iabl es X and Y a e said to be positively co elated and this indicates that Y tends to inc ease when X does; when the sign is negative, the va iables a e said to be negatively co elated and this indicates that Y tends to dec ease when X inc ea ses; and when the sign is 0, the va iables a e said to be unco elated. Figu e 3 6.1 shows some examples of data pai s and thei co elation. Figu e 36.1

366 PROPERTIES OF EXPECTATION P actice P oblems P oblem 36.1 If X and Y a e independent and identically dist ibuted with mean an d va iance 2 , nd E[(X Y )2 ]. Problem 36.2 Two card are drawn without replaceme nt from a pack of card . The random variable X mea ure the number of heart card drawn, and the random variable Y mea ure the number of club card drawn. Find the covariance and correlation of X and Y. Problem 36.3 Suppo e the joint pdf o f X and Y i fXY (x, y) = 1 0 < x < 1, x < y < x + 1 0 otherwi e Compute the covariance and correlation of X and Y.. Problem 36.4 Let X and Z be independent random variable with X uniformly di tributed on (1, 1) and Z uniform ly di tributed on (0, 0.1). Let Y = X 2 + Z. Then X and Y are dependent. (a) Fin d the joint pdf of X and Y. (b) Find the covariance and the correlation of X and Y. Problem 36.5 Let the random variable be unifo mly dist ibuted on [0, 2]. Cons ider the random variables X = cos and Y = sin . Show that Cov(X, Y ) = 0 even tho ugh X and Y a e dependent. This means that the e is a weak elationship between X and Y. P oblem 36.6 If X1 , X2 , X3 , X4 a e (pai wise) unco elated andom va iables each having mean 0 and va iance 1, compute the co elations of (a) X1 + X2 and X2 + X3 . (b) X1 + X2 and X3 + X4 .

36 COVARIANCE, VARIANCE OF SUMS, AND CORRELATIONS 367 P oblem 36.7 Let X be the numbe of 1s and Y the numbe of 2s that occu in n oll s of a fai die. Compute Cov(X, Y ). P oblem 36.8 Let X be unifo mly dist ibuted on [1, 1] and Y = X 2 . Show that X and Y a e unco elated even though Y depends functionally on X (the st ongest fo m of dependence). P oblem 36.9 Let X and Y be continuous andom va iables with joint pdf fXY (x, y) = Find Cov(X, Y ) and (X , Y ). P oblem 36.10 Suppose that X and Y a e andom va iables with Cov(X, Y ) = 3. Find Cov(2X 5, 4Y + 2). P oblem 36.11 An insu ance policy pays a total medic al bene t consisting of two pa ts fo each claim. Let X ep esent the pa t of the bene t that is paid to the su geon, and let Y ep esent the pa t that is paid to t he hospital. The va iance of X is 5000, the va iance of Y is 10,000, and the va iance of the total bene t, X + Y, is 17,000. Due to inc easing medical costs, the company that issues the policy decides to inc ease X by a at amount of 100 pe cl aim and to inc ease Y by 10% pe claim. Calculate the va iance of the total bene t afte these evisions have been made. P oblem 36.12 The p o t fo a new p oduct i s given by Z = 3X Y 5. X and Y a e independent andom va iables with Va (X) = 1 and Va (Y) = 2. What is the va iance of Z? 3x 0 y x 1 0 othe wise

368 PROPERTIES OF EXPECTATION P oblem 36.13 A company has two elect ic gene ato s. The time until failu e fo each gene ato follows an exponential dist ibution with mean 10. The company wil l begin using the second gene ato immediately afte the st one fails. What is t he va iance of the total time that the gene ato s p oduce elect icity? P oblem 3 6.14 A joint density function is given by fXY (x, y) = Find Cov(X, Y ) P oblem 3 6.15 Let X and Y be continuous andom va iables with joint density function fXY (x, y) = Find Cov(X, Y ) P oblem 36.16 Let X and Y denote the values of two stoc ks at the end of a ve yea pe iod. X is unifo mly dist ibuted on the inte val (0, 12) . Given X = x, Y is unifo mly dist ibuted on the inte val (0, x). Dete mine Cov(X, Y ) acco ding to this model. P oblem 36.17 Let X denote the size of a su gical claim and let Y denote the size of the associated hospital claim. An actu a y is using a model in which E(X) = 5, E(X 2 ) = 27.4, E(Y ) = 7, E(Y 2 ) = 51. 4, and V a (X + Y ) = 8. Let C1 = X +Y denote the size of the combined claims be fo e the application of a 20% su cha ge on the hospital po tion of the claim, an d let C2 denote the size of the combined claims afte the application of that su cha ge. Calculate Cov(C1 , C2 ). P oblem 36.18 Claims led unde auto insu ance p olicies follow a no mal dist ibution with mean 19,400 and standa d deviation 5,0 00. What is the p obability that the ave age of 25 andomly selected claims exce eds 20,000 ? 8 xy 3 kx 0 < x, y < 1 0 othe wise 0 0 x 1, x y 2x othe wise

369 (a) W ite down the joint pdf fXY (x, y). (b) Let Z = X + Y. Find the pdf fZ (a). Simplify as much as possible. (c) Find the expectation E(X) and va iance Va (X) . Repeat fo Y. (d) Compute the expectation E(Z) and the va iance Va (Z). P oble m 36.20 Let X and Y be two andom va iables with joint pdf fXY (x, y) = 1 2 0 x > 0, y > 0, x + y < 2 othe wise (a) Let Z = X + Y. Find the pdf of Z. (b) Find the pdf of X and that of Y. (c) F ind the expectation and va iance of X. (d) Find the cova iance Cov(X, Y ). P obl em 36.21 Let X and Y be disc ete andom va iables with joint dist ibution de ned b y the following table Y\ X 0 1 2 pY (y) 2 0.05 0.40 0.05 0.50 3 0.05 0 0.15 0.20 4 0.15 0 0.10 0.25 5 0.05 0 0 0.05 pY (y) 0.30 0.40 0.30 1 Fo this joint dist ibution, E(X) = 2.85, E(Y ) = 1. Calculate Cov(X, Y). P oble m 36.22 Let X and Y be two andom va iables with joint density function fXY (x, y) = Find Va (X). 1 0 < y < 1 x , 1 < x 1 0 otherwise
36 COVARIANCE, VARIANCE OF SUMS, AND CORRELATIONS P oblem 36.19 Let X and Y be t wo independent andom va iables with densities fX (x) = fY (y) = 1 0<x<1 0 othe wise 0<y<2 0 othe wise 1 2

370 PROPERTIES OF EXPECTATION Problem 36.23 Let X1 , X2 , X3 be uniform random variables on the interval (0, 1 ) with Cov(Xi , Xj ) = 1 for i, j {1, 2, 3}, i = j. Calculate the variance of X1 + 2X2 X3 . 24 Problem 36.24 Let X and X be discrete random variables with joint probability function pXY (x, y) given by the following table: X\ Y 0 1 2 pY (y) Find the variance of Y X. Problem 36.25 Let X and Y be two independent identica lly distributed normal random variables with mean 1 and variance 1. Find c so th at E[c X Y ] = 1. Problem 36.26 Let X, Y and Z be random variables with means 1 ,2 and 3, respectively, and variances 4,5, and 9, respectively. Also, Cov(X, Y ) = 2, Cov(X, Z) = 3, and Cov(Y, Z) = 1. What are the mean and variance, respecti vely, of the random variable W = 3X + 2Y Z? Problem 36.27 Let X1 , X2 , and X3 b e independent random variables each with mean 0 and variance 1. Let X = 2X1 X3 a nd Y = 2X2 + X3 . Find (X, Y ). P oblem 36.28 The coe cient of co elation between andom va iables X and Y is 1 , and 3 2 2 X = a, Y = 4a. The random variable Z i de ned to be Z = 3X 4Y, and 2 it i found that Z = 114. Find a. Problem 36.29 Given n independent random variable X1 , X2 , , Xn each having the same variance 2 . De ne U = 2X1 + X2 + + Xn1 and V = X2 + X3 + + Xn1 + 2Xn . Find (U, V ). 0 20 0.60 1 pX (x) 0.20 0.20 0.20 0.60 0 0.20 0.40 1

36 COVARIANCE, VARIANCE OF SUMS, AND CORRELATIONS 371 P oblem 36.30 The following table gives the joint p obability dist ibution fo t he numbe s of washe s (X) and d ye s (Y) sold by an appliance sto e salespe son in a day. X\Y 0 1 2 0 0.25 0.12 0.03 1 0.08 0.20 0.07 2 0.05 0.10 0.10 (a) Give the ma ginal dist ibutions of numbe s of washe s and d ye s sold pe da y. (b) Give the Expected numbe s of Washe s and D ye s sold in a day. (c) Give t he cova iance between the numbe of washe s and d ye s sold pe day. (d) If the salespe son makes a commission of $100 pe washe and $75 pe d ye , give the av e age daily commission. P oblem 36.31 In a la ge class, on exam 1, the mean and standa d deviation of sco es we e 72 and 15, espectively. Fo exam 2, the mean and standa d deviation we e 68 and 20, espectively. The cova iance of the exam sco es was 120. Give the mean and standa d deviation of the sum of the two exam sco es. Assume all students took both exams

372 PROPERTIES OF EXPECTATION 37 Conditional Expectation Since conditional p obability measu es a e p obabilitiy measu es (that is, they possess all of the p ope ties of unconditional p obability measu es), conditiona l expectations inhe it all of the p ope ties of egula expectations. Let X and Y be andom va iables. We de ne conditional expectation of X given that Y = y by E (X Y = y} = x xP (X = x Y = y) xpX Y (x y) x = where pX Y is the conditional probability mass function of X, given that Y = y w hich is given by pX Y (x y) = P (X = x Y = y) = In the continuous case we have p(x, y) . pY (y) E(X Y = y) = xfX Y (x y)dx fXY (x, y) . fY (y) where fX Y (x y) = Example 37.1 Suppose X and Y are discrete random variables with values 1, 2, 3, 4 and joint p.m.f. given by 1 16 if x = y 2 if x < y f (x, y) = 16 0 if x > y fo r x, y = 1, 2, 3, 4. (a) Find the joint probability distribution of X and Y. (b) Find the conditional expectation of Y given that X = 3. Solution. (a) The joint probability distribution is given in tabular form

37 CONDITIONAL EXPECTATION X\Y 1 2 3 4 pY (y) (b) We have 4 373 3 2 16 2 16 1 16 1 1 16 2 2 16 1 16 4 2 16 2 16 2 16 1 16 7 16 pX (x) 7 16 5 16 3 16 1 16 0 0 0 1 16 0 0 3 16 0 5 16 1 E(Y X = 3) = y=1 ypY X (y 3) pXY (3, 1) 2pXY (3, 2) 3pXY (3, 3) 4pXY (3, 4) + + + pX (3) pX (3) pX (3) pX (3) 1 2 11 =3 + 4 = 3 3 3 = Example 37.2 Suppose that the joint density of X and Y is given by e y ey , fXY (x, y) = y Compute E(X Y = y). Solution. The conditional density is found as follows fX Y (x y) = fXY (x, y) fY (y) fXY (x, y) = f (x, y) dx XY = = (1/y)e y ey x (1/y)e y ey dx 0 x y x x x, y > 0. (1/y)e x (1/y)e y dx 0 1 x = e y y
374 Hence, PROPERTIES OF EXPECTATION E(X Y = y) = 0 x x x e y dx = xe y y 0 0 e y dx x = xe x y + ye x y =y 0 Example 37.3 Let Y be a random variable with a density fY given by fY (y) = 1 y 0 y>1 otherwise where > 1. Given Y = y, let X be r ndom v ri ble which is Uniformly distribute d on (0, y). ( ) Find the m rgin l distribution of X. (b) C lcul te E(Y X = x) for every x > 0. Solution. The joint density function is given by fXY (x, y) = 1 y +1 0 0 < x < y, y > 1 otherwise ( ) Observe th t X only t kes positive v lues, thus fX (x) = 0, x 0. For 0 < x < 1 we h ve fX (x) = fXY (x, y)dy = 1
fXY (x, y)dy = 1 For x 1 we have fX (x) = fXY (x, y) y = x fXY (x, y) y = 1 x (b) For 0 < x < 1 we h ve fY X (y x) = Hence, E(Y X = x) = 1 fXY (x, y) = +1 , y > 1. fX (x) y y dy = y +1 1 dy = . y 1

37 CONDITIONAL EXPECTATION If x 1 then fY X (y x) = Hence, 375 fXY (x, y) x = +1 , y > x. fX (x) y E(Y X = x) = x y x x dy = +1 y 1 Notice th t if X nd Y re independent then pX Y (x y) = p(x) so that E(X Y = y) = E(X). Now, for any function g(x), the conditional expected value of g given Y = y is, in the continuous case, E(g(X) Y = y) = g(x)fX Y (x y)dx if the integral exists. For the discrete case, we have a sum instead of an integ ral. That is, the conditional expectation of g given Y = y is E(g(X) Y = y) = x g(x)pX Y (x y). The proof of this result is identical to the unconditional case. Next, let X (y) = E(X Y = y) denote the function of the random variable Y whose value at Y = y i s E(X Y = y). Clearly, X (y) is a random variable. We denote this random variable by E(X Y ). The expectation of this random variable is just the expectation of X as shown in the following theorem. Theorem 37.1 (Double Expectation Property) E(X) = E(E(X Y ))
376 PROPERTIES OF EXPECTATION Proof. We give a proof in the case X and Y are continuous random variables. E(E(X Y )) = E(X Y = y)fY (y)dy = xfX Y (x y)dx fY (y)dy xfX Y (x y)fY (y)dxdy = = x fXY (x, y)dydx = xfX (x)dx = E(X) Computing Probabilities by Conditioning Suppose we want to know the probability of some event, A. Suppose also that knowing Y gives us some useful information a bout whether or not A occurred. De ne an indicator random variable X= Then P (A) = E(X) and for any random variable Y E(X Y = y) = P (A Y = y). Thus, by the doubl e expectation property we have P (A) =E(X) = y 1 if A occurs 0 if A does not occur E(X Y = y)P (Y = y) = y P (A Y = y)pY (y) in the discrete case and P (A) = P (A Y = y)fY (y)dy
37 CONDITIONAL EXPECTATION in the continuous case. 377 The Conditional Variance Next, we introduce the concept of conditional variance. Just as we have de ned the conditional expectation of X given that Y = y, we can de ne the conditional variance of X given Y as follows Var(X Y = y) = E[(X E(X Y ) )2 Y = y]. Note that the conditional variance is a random variable since it is a function of Y. Proposition 37.1 Let X and Y be random variables. Then (a) Var( X Y ) = E(X 2 Y ) [E(X Y )]2 (b) E(Var(X Y )) = E[E(X 2 Y ) (E(X Y ))2 ] = E(X 2 ) E[(E(X Y ))2 ]. (c) Var(E(X Y )) = E[(E(X Y ))2 ] (E(X))2 . (d) Var(X) = E[ Var(X Y )] + Var(E(X Y )). Proof. (a) We have Var(X Y ) =E[(X E(X Y ))2 Y = y] =E[(X 2 2XE(X Y ) + (E(X Y ))2 Y ] =E(X 2 Y ) 2E(X Y )E(X Y ) + (E(X Y ))2 =E( X 2 Y ) [E(X Y )]2 (b) Taking E of both sides of the result in (a) we nd E(Var(X Y )) = E[E(X 2 Y ) (E(X Y ))2 ] = E(X 2 ) E[(E(X Y ))2 ]. (c) Since E(E(X Y )) = E(X) we have Var(E(X Y )) = E[(E(X Y ))2 ] (E(X))2 . (d) The result follows b y adding the two equations in (b) and (c) Conditional Expectation and Prediction One of the most important uses of conditional expectation is in estimation
378 PROPERTIES OF EXPECTATION theory. Let us begin this discussion by asking: What constitutes a good estimato r? An obvious answer is that the estimate be close to the true value. Suppose th at we are in a situation where the value of a random variable is observed and th en, based on the observed, an attempt is made to predict the value of a second r andom variable Y. Let g(X) denote the predictor, that is, if X is observed to be equal to x, then g(x) is our prediction for the value of Y. So the question is of choosing g in such a way g(X) is close to Y. One possible criterion for close ness is to choose g so as to minimize E[(Y g(X))2 ]. Such a minimizer will be ca lled minimum mean square estimate (MMSE) of Y given X. The following theorem sho ws that the MMSE of Y given X is just the conditional expectation E(Y X). Theor em 37.2 min E[(Y g(X))2 ] = E(Y E(Y X)). g Proof. We have E[(Y g(X))2 ] =E[(Y E(Y X) + E(Y X) g(X))2 ] =E[(Y E(Y X))2 ] + E[(E(Y X) g(X))2 ] +2E[(Y E(Y X))(E(Y X) g(X))] Using the fact that the exp ression h(X) = E(Y X) g(X) is a function of X and thus can be treated as a cons tant we have E[(Y E(Y X))h(X)] =E[E[(Y E(Y X))h(X) X]] =E[(h(X)E[Y E(Y X) X]] =E[h(X)[E(Y X) E(Y X)]] = 0 for all functions g. Thus, E[(Y g(X))2 ] = E[(Y E (Y X))2 ] + E[(E(Y X) g(X))2 ]. The rst term on the right of the previous equat ion is not a function of g. Thus, the right hand side expression is minimized wh en g(X) = E(Y X)
37 CONDITIONAL EXPECTATION 379 Practice Problems Problem 37.1 Suppose that X and Y have joint distribution fXY (x, y) = Find E(X Y ) and E(Y X). Problem 37.2 Suppose that X and Y have joint distribution fXY ( x, y) = 3y 2 x3 8xy 0 < x < y < 1 0 otherwise 0 0<y<x<1 otherwise
Find E(X), E(X 2 ), V ar(X), E(Y X), V ar(Y X), E[V ar(Y X)], V ar[E(Y X)], and V ar(Y ). Problem 37.3 Let X and Y be independent exponentially distributed random variables with parameters and respective y. Using conditioning, nd P (X > Y ). Prob em 37.4 Let X be uniform y distributed on [0, 1]. Find E(X X > 0.5). P roblem 37.5 Let X and Y be discrete random variables with conditional density fu nction y=1 0.2 0.3 y=2 fY X (y 2) = y=3 0.5 0 otherwise Compute E(Y X = 2). oblem 37.6 Suppose that X and Y have joint distribution fXY (x, y) = Find E(Y X ). 21 2 xy 4 0 x2 < y < 1 otherwise
380 PROPERTIES OF EXPECTATION Problem 37.7 Suppose that X and Y have joint distribution fXY (x, y) = Find E(Y ) in two ways. Problem 37.8 Suppose that E(X Y ) = 18 3 Y and E(Y X) = 10 1 X. Find E(X) and 5 3 E(Y ). Problem 37.9 Let X be an exponential random variable wi th = 5 and Y a uniform y distributed random variab e on (3, X). Find E(Y ). Prob em 37.10 The number of peop e who pass by a store during unch time (say form 12 :00 to 1:00 pm) is a Poisson random variab e with parameter = 100. Assume that e ach person may enter the store, independent y of the other peop e, with a given probabi ity p = 0.15. What is the expected number of peop e who enter the store during unch time? Prob em 37.11 Let X and Y be discrete random variab es with j oint probabi ity mass function de ned by the fo owing tab e X\Y 1 2 3 pY (y) 1 1/ 9 1/3 1/9 5/9 2 1/9 0 1/18 1/6 3 0 1/6 1/9 5/18 pX (x) 2/9 1/2 5/18 1 21 2 xy 4 0 x2 < y < 1 otherwise Compute E(X Y = i) for i = 1, 2, 3. Are X and Y independent? Problem 37.12 A dia gnostic test for the presence of a disease has two possible outcomes: 1 for dise ase present and 0 for disease not present. Let X denote the disease
37 CONDITIONAL EXPECTATION 381 state of a patient, and let Y denote the outcome of the diagnostic test. The joi nt probability function of X and Y is given by: P (X P (X P (X P (X Calculate V ar(Y X = 1). Problem 37.13 The stock prices of two companies at the end of any given year are modeled with random variables X and Y that follow a distribution with joint density function 2x 0 < x < 1, x < y < x + 1 fXY (x, y) = 0 otherwise What is the conditional variance of Y given that X = x? Problem 37.14 An actuar y determines that the annual numbers of tornadoes in counties P and Q are jointl y distributed as follows: X\Y 0 1 2 3 pY (y) 0 0.12 0.06 0.05 0.02 0.25 1 0.13 0 .15 0.12 0.03 0.43 2 0.05 0.15 0.10 0.02 0.32 PX (x) 0.30 0.36 0.27 0.07 1 = 0, Y = 1, Y = 0, Y = 1, Y = 0) =0.800 = 0) =0.050 = 1) =0.025 = 1) =0.125 where X is the number of tornadoes in county Q and Y that of county P. Calculate the conditional variance of the annual number of tornadoes in county Q, given t hat there are no tornadoes in county P. Problem 37.15 Let X be a random variable with mean 3 and variance 2, and let Y be a random variable such that for every x, the conditional distribution of Y given X = x has a mean of x and a variance of x2 . What is the variance of the marginal distribution of Y ?
382 PROPERTIES OF EXPECTATION Problem 37.16 Let X and Y be two continuous random variables with joint density function fXY (x, y) = For 0 < x < 1, nd Var(Y X = x). Problem 37.17 The number o f stops X in a day for a delivery truck driver is Poisson with mean . Conditiona on their being X = x stops, the expected distance driven by the driver Y is Nor ma with a mean of x miles, nd st nd rd devi tion of x miles. Give the mean and variance of the num ers of miles she drives per day. 2 0<x<y<1 0 otherwise
38 MOMENT GENERATING FUNCTIONS 383 38 Moment Generating Functions The moment generating function of a random varia le X, denoted y MX (t), is de ne d as MX (t) = E[etX ] provided that the expectation exists for t in some neigh o rhood of 0. For a discrete random varia le with a pmf p(x) we have MX (t) = x etx p(x) and for a continuous random varia le with pdf f, MX (t) = etx f (x)dx. Example 38.1 Let X e a discrete random varia le with pmf given y the following ta le x 1 2 3 4 5 p(x) 0.15 0.20 0.40 0.15 0.10 Find MX (t). Solution. We have MX (t) = 0.15et + 0.20e2t + 0.40e3t + 0.15e4t + 0.10e5t Example 38.2 Let X e th e uniform random varia le on the interval [a, ]. Find MX (t). Solution. We have MX (t) = a
etx 1 dx = [et eta ]
a t( a)
As the name suggests, the moment generating function can e used to generate mom ents E(X n ) for n = 1, 2, . Our rst result shows how to use the moment generatin g function to calculate moments.
384 Proposition 38.1 PROPERTIES OF EXPECTATION n E(X n ) = MX (0) where n MX (0) = dn MX (t) t=0 dtn Proof. We prove the result for a continuous random variable X with pdf f. The di screte case is shown similarly. In what follows we always assume that we can di er entiate under the integral sign. This interchangeability of di erentiation and exp ectation is not very limiting, since all of the distributions we will consider e njoy this property. We have d d MX (t) = dt dt = etx f (x)dx = d tx e f (x)dx dt xetx f (x)dx = E[XetX ] Hence, d MX (t) t=0 = E[XetX ] t=0 = E(X). dt By induction on n we nd dn MX (t) t=0 = E[X n etX ] t=0 = E(X n ) dtn We next compute MX (t) for some common di stributions. Example 38.3 Let X be a binomial random variable with parameters n and p. Find the expected value and the variance of X using moment generating fun ctions. Solution. We can write n MX (t) =E(etX ) = k=0 n etk C(n, k)pk (1 p)nk = k=0 C(n, k)(pet )k (1 p)nk = (pet + 1 p)n
38 MOMENT GENERATING FUNCTIONS Di erentiating yields d MX (t) = npet (pet + 1 p)n1 dt Thus d MX (t) t=0 = np. dt To nd E(X 2 ), we di erentiate a second time to obta in E(X) = d2 MX (t) = n(n 1)p2 e2t (pet + 1 p)n2 + npet (pet + 1 p)n1 . dt2 Evalua ting at t = 0 we nd E(X 2 ) = MX (0) = n(n 1)p2 + np. Observe that this implies t he variance of X is V ar(X) = E(X 2 ) (E(X))2 = n(n 1)p2 + np n2 p2 = np(1 p) 385 Example 38.4 Let X be a Poisson random variable with parameter . Find the expecte d va ue and the variance of X using moment generating functions. So ution. We ca n write MX (t) =E(e ) = n=0 tX etn n etn e n = e n! n! n=0 =e n=0 (et )n t t = e ee = e(e 1) n! Di erentiating for the t rst time we nd MX (t) = et e(e 1) . Thus, E(X) = MX (0) = .
386 Di erentiating a second time we PROPERTIES OF EXPECTATION
nd
MX (t) = (et )2 e(e 1) + et e(e 1) . Hence, E(X 2 ) = MX (0) = 2 + . The variance i n V ar(X) = E(X 2 ) (E(X))2 = Examp e 38.5 Let X be an exponentia random variab e with parameter . Find the expected va ue and the variance of X using moment ge nerating functions. So ution. We can write MX (t) = E(etX ) = tx e ex dx 0 t t = (t)x e dx 0 = t where t < . Di erentiation twice yie ds MX (t) = Hence, E(X) = MX (0) = The varianc e of X is given by V ar(X) = E(X 2 ) (E(X))2 = 1 2 1 (t)2 and MX (t) = 2 . (t)3 and E(X 2 ) = MX (0) = 2 . 2 Moment generating functions are a so usefu in estab ishing the distribution of sums of independent random variab es. To see this, the fo owing two observation s are usefu . Let X be a random variab e, and et a and b be nite constants. Then , MaX+b (t) =E[et(aX+b) ] = E[ebt e(at)X ] =ebt E[e(at)X ] = ebt MX (at)
38 MOMENT GENERATING FUNCTIONS 387
Examp e 38.6 Let X be a norma random variab e with parameters and 2 . Find the expected value and the variance of X u ing moment generating function . Solution . Fir t we nd the moment of a tandard normal random variable with parameter 0 a nd 1. We can write z2 (z 2 2tz) 1 1 etz e 2 dz = exp dz MZ (t) =E(etZ ) = 2 2 2 (zt) 2 = ex + dz = e e 2 dz = e 2 2 2 2 2
Now, since X = + Z we have MX (t) =E(etX ) = E(et+tZ ) = E(et etZ ) = et E(etZ ) t2 + t =et MZ (t) = et e 2 = exp 2 By di erentiation we obtain MX (t) = ( + t 2 )exp d MX (t) = ( + t 2 )2 exp and thu E(X) = MX (0) = and E(X 2 ) = MX (0) = 2 + 2 The variance of X i V ar(X) = E(X 2 ) (E(X))2 = 2 Next, uppo e X1 , X2 , , XN are independent random variables. Then, the moment generating function of Y = X1 + + XN is MY (t) =E(et(X1 +X2 ++Xn ) ) = E(eX1 t eXN t ) N N 2 t2 + t 2 2 t2 + t 2 2 t2 + t + 2 exp 2 = k=1 E(e Xk t )= k=1 MXk (t)
388 PROPERTIES OF EXPECTATION where the next to la t equality follow from Propo ition 35.5. Another important property i that the moment generating function uniquely determine the di trib ution. That i , if random variable X and Y both have moment generating function MX (t) and MY (t) that exi t in ome neighborhood of zero and if MX (t) = MY ( t) for all t in thi neighborhood, then X and Y have the ame di tribution . The general proof of thi i an inver ion problem involving Laplace tran form theor y and i omitted. However, We will prove the claim here in a impli ed etting. Su ppo e X and Y are two random variable with common range {0, 1, 2, , n}. Moreove r, suppose that both variables have the same moment generating function. That is , n n etx pX (x) = x=0 y=0 ety pY (y). For simplicity, let s = et and ci = pX (i) pY (i) for i = 0, 1, , n. Then n n 0= x=0 n e pX (x) y=0 n tx ety pY (y) sy pY (y) y=0 n 0= x=0 n sx pX (x) si pX (i) i=0 n i=0 0= 0= i=0 n si pY (i) si [pX (i) pY (i)] ci si , s > 0. i=0 0= The above is simply a polynomial in s with coe cients c0 , c1 , , cn . The only wa y it can be zero for all values of s is if c0 = c1 = = cn = 0. That is pX (i) = pY (i), i = 0, 1, 2, , n. So probability mass functions for X and Y are exactly the same.
38 MOMENT GENERATING FUNCTIONS 389
Example 38.7 If X and Y are independent binomial random variables with parameter s (n, p) and (m, p), respectively, what is the pmf of X + Y ? Solution. We have MX+Y (t) =MX (t)MY (t) =(pet + 1 p)n (pet + 1 p)m =(pet + 1 p)n+m . Since (pet + 1p)n+m is the moment generating function of a binomial random variable having par ameters m+n and p, X +Y is a binomial random variable with this same pmf Example 38.8 If X and Y are independent Poisson random variables with parameters 1 and 2 , respective y, what is the pmf of X + Y ? So ution. We have MX+Y (t) =MX (t)MY (t) =e1 (e 1) e2 (e 1) =e(1 +2 )(e 1) . Since e(1 +2 )(e 1) is the moment generat ion of a Poisson random variab e having parameter 1 + 2 , X + Y is a Poisson rando m variab e with this same pmf Examp e 38.9 2 If X and Y are independent norma r andom variab es with parameters (1 , 1 ) 2 and (2 , 2 ), re pectively, what i the d i tribution of X + Y ? Solution. We have MX+Y (t) =MX (t)MY (t) 2 2 1 t2 2 t2 =exp + 1 t exp + 2 t 2 2 2 2 (1 + 2 )t2 =exp + (1 + 2 )t 2 t t t t
390 PROPERTIES OF EXPECTATION which i the moment generating function of a normal random variable with 2 2 mea n 1 + 2 and variance 1 + 2 . Becau e the moment generating function uniquley determi ne the di tribution then X +Y i a normal random variable with the ame di trib ution Example 38.10 An in urance company in ure two type of car , economy car and luxury car . The damage claim re ulting from an accident involving an econo my car ha normal N (7, 1) di tribution, the claim from a luxury car accident ha normal N (20, 6) di tribution. Suppo e the company receive three claim from economy car accident and one claim from a luxury car accident. A uming that th e e four claim are mutually independent, what i the probability that the total claim amount from the three economy car accident exceed the claim amount from the luxury car accident? Solution. Let X1 , X2 , X3 denote the claim amount fr om the three economy car , and X4 the claim from the luxury car. Then we need to compute P (X1 +X2 +X3 > X4 ), which i the ame a P (X1 + X2 + X3 X4 > 0). Now , ince the Xi are independent and normal with di tribution N (7, 1) (for i = 1, 2, 3) and N (20, 6) for i = 4, the linear combination X = X1 +X2 +X3 X4 ha no rmal di tribution with parameter = 7+7+720 = 1 and 2 = 1+1+1+6 = 9. Thu , the pr obability we want i P (X > 0) =P 01 X 1 > 9 9 =P (Z > 0.33) = 1 P (Z 0.33) =P 33) 0.6293 Joint Moment Generating Function For any random variable X1 , X2 , , Xn , the joint moment generating function is de ned by M (t1 , t2 , , tn ) = E(et1 X1 +t2 X 2 ++tn Xn ). Example 38.11 Let X and Y be two independent normal random variables w ith parameters
38 MOMENT GENERATING FUNCTIONS 391 2 2 (1 , 1 ) and (2 , 2 ) re pectively. Find the joint moment generating function of X + Y and X Y. Solution. The joint moment generating function i M (t1 , t2 ) =E(et1 (X+Y )+t2 (XY ) ) = E(e(t1 +t2 )X+(t1 t2 )Y ) =E(e(t1 +t2 )X )E(e(t1 t2 )Y ) = MX (t1 + t2 )M Y (t1 t2 ) =e(t1 +t2 )1 + 2 (t1 +t2 ) 1 2 2 1 e(t1 t2 )2 + 2 (t1 t2 ) 2 2 2 1 2 2 1 2 2 2 2 2 2
=e(t1 +t2 )1 +(t1 t2 )2 + 2 (t1 +t2 )1 + 2 (t1 +t2 )2 +t1 t2 (1 2 ) Example 38.12 L and Y be two random variable with joint di tribution function fXY (x, y) = exy x > 0, y > 0 0 otherwi e 1
1 1 . 1 t1 1 t2 1 (1 t1 )2 (1 t2 )2 (0,0) =1 E(X) = E(Y ) = and = (0,0) 1 (1 t1 )2 (1 t2 ) 1 (1 t1 )(1 t2 )2 =1 (0,0) = (0,0) =1 (0,0) Cov(X, Y ) = E(XY ) E(X)E(Y ) = 0
Find E(XY ), E(X), E(Y fX (x)fY (y) o that X on i given by M (t1 , = 2 M (t1 , t2 ) t2 (0,0)
) and Cov(X, Y ). Solution. We note r t that fXY (x, y) = and Y are independent. Thu , the moment generating functi t2 ) = E(et1 X+t2 Y ) = E(et1 X )E(et2 Y ) = Thu , E(XY ) t1 M (t1 , t2 ) t1 M (t1 , t2 ) t2 =
392 PROPERTIES OF EXPECTATION Practice Problem Problem 38.1 Let X be a di crete random variable with range {1, 2, , n} so that its pmf 1 is given by pX (j) = n for 1 j n. Find E(X) and V ar(X) using moment g enetating functions. Problem 38.2 Let X be a geometric distribution function wit h pX (n) = p(1 p)n1 . Find the expected value and the variance of X using moment generating functions. Problem 38.3 The following problem exhibits a random varia ble with no moment generating function. Let X be a random variable with pmf give n by pX (n) = 6 2 n2 , n = 1, 2, 3, . Show that MX (t) does not exist in any neighborhood of 0. Problem 38.4 Let X be a gamma random variable with parameters nd . Find the expected va ue and the var iance of X using moment generating functions. Prob em 38.5 Show that the sum of n independent y exponentia random variab e each with paramter is a gamma random variab e with parameters n and . Prob em 38.6 Let X be a random variab e with pd f given by f (x) = Find MX (t). Prob em 38.7 Let X be an exponentia random vari ab e with paramter . Find the moment generating function of Y = 3X 2. 1 , (1 + x2 ) < x < .
38 MOMENT GENERATING 393
Problem 38.8 Identify the random variable whose moment generating function is gi ven by MX (t) = 3 t 1 e + 4 4 15 . Problem 38.9 Identify the random variable whose moment generating function is gi ven by MY (t) = e2t 3 3t 1 e + 4 4 15 . Problem 38.10 X and Y a e independent andom va iables with common moment gene a ting t2 function M (t) = e 2 . Let W = X + Y and Z = X Y. Dete mine the joint mo ment gene ating function, M (t1 , t2 ) of W and Z. P oblem 38.11 An actua y dete mines that the claim size fo a ce tain class of accidents is a andom va iable , X, with moment gene ating function MX (t) = 1 (1 2500t)4 Dete mine the standa d deviation of the claim size fo this class of accidents. P oblem 38.12 A company insu es homes in th ee cities, J, K, and L . Since su cien t distance sepa ates the cities, it is easonable to assume that the losses occu ing in these cities a e independent. The moment gene ating functions fo the l oss dist ibutions of the cities a e: MJ (t) =(1 2t)3 MK (t) =(1 2t)2.5 MJ (t) =(1 2t)4.5 Let X ep esent the combined losses f om the th ee cities. Calculate E(X 3 ).

UNCTIONS

394 PROPERTIES OF EXPECTATION P oblem 38.13 Let X1 , X2 , X3 be disc ete andom va iables with common p obabil ity mass function 1 x=0 3 2 x=1 p(x) = 3 0 othe wise Dete mine the moment gene a ting function M (t), of Y = X1 X2 X3 . P oblem 38.14 Two inst uments a e used to measu e the height, h, of a towe . The e o made by the less accu ate inst ume nt is no mally dist ibuted with mean 0 and standa d deviation 0.0056h. The e o made by the mo e accu ate inst ument is no mally dist ibuted with mean 0 and st anda d deviation 0.0044h. Assuming the two measu ements a e independent andom v a iables, what is the p obability that thei ave age value is within 0.005h of t he height of the towe ? P oblem 38.15 Let X1 , X2 , , Xn be independent geometri c random variables each with parameter p. De ne Y = X1 + X2 + Xn . (a) Find the mo ment generating function of Xi , 1 i n. (b) Find the moment generating function of a negative binomial random variable with parameters (n, p). (c) Show that Y d e ned above is a negative binomial random variable with parameters (n, p). Problem 38.16 Assume the math scores on the SAT test are normally distributed with mean 500 and standard deviation 60, and the verbal scores are normally distributed w ith mean 450 and standard deviation 80. If two students who took both tests are chosen at random, what is the probability that the rst students math score exceeds the second students verbal score? Problem 38.17 Suppose a random variable X has moment generating function MX (t) = Find the variance of X. 2 + et 3 9

38 MOMENT GENERATING FUNCTIONS Problem 38.18 Let X be a random variable with den sity function f (x) = (k + 1)x2 0 < x < 1 0 otherwise 395 Find the moment generating function of X Problem 38.19 If the moment generating function for the random variable X is MX (t) = nd E[(X 2)3 ]. 1 , t+1 Problem 38.20 Suppose that X is a random variable with moment generating functio n (tj1) MX (t) = e j! . Find P (X = 2). j=0 Problem 38.21 If X has a standard nor mal distribution and Y = eX , what is the k th moment of Y ? Problem 38.22 The r andom variable X has an exponential distribution with parameter b. It is found t hat MX (b2 ) = 0.2. Find b. Problem 38.23 Let X1 and X2 be two random variables w ith joint density function fX1 X1 (x1 , x2 ) = 1 0 < x1 < 1, 0 < x2 < 1 0 otherw ise Find the moment generating function M (t1 , t2 ). Problem 38.24 The moment gener ating function for the joint distribution of random vari2 1 ables X and Y is M ( t1 , t2 ) = 3(1t2 ) + 2 et1 (2t2 ) , t2 < 1. Find Var(X). 3 Problem 38.25 Let X an d Y be two independent random variables with moment generating functions
396 MX (t) = et 2 +2t PROPERTIES OF EXPECTATION and MY (t) = e3t 2 +t Determine the moment generating function of X + 2Y. Problem 38.26 Let X1 and X2 be random variables with joint moment generating function M (t1 , t2 ) = 0.3 + 0 .1et1 + 0.2et2 + 0.4et1 +t2 What is E(2X1 X2 )? Problem 38.27 Suppose X and Y ar e random variables whose joint distribution has moment generating function MXY ( t1 , t2 ) = 1 t1 3 t2 3 e + e + 4 8 8 10 for all t1 , t2 . Find the covariance between X and Y. Problem 38.28 Independent random variables X, Y and Z are identically distributed. Let W = X +Y. The mome nt generating function of W is MW (t) = (0.7+0.3et )6 . Find the moment generati ng function of V = X + Y + Z. Problem 38.29 The birth weight of males is normall y distributed with mean 6 pounds, 10 ounces, standard deviation 1 pound. For fem ales, the mean weight is 7 pounds, 2 ounces with standard deviation 12 ounces. G iven two independent male/female births, nd the probability that the baby boy out weighs the baby girl. Problem 38.30 The value of a piece of factory equipment af ter three years of use is 100(0.5)X where X is a random variable having moment g enerating function MX (t) = 1 12t for t < 1 . 2 Calculate the expected value of this piece of equipment after three years of use .
Limit Theorems Limit theorems are considered among the important results in probability theory. In this chapter, we consider two types of limit theorems. The rst type is known as the law of large numbers. The law of large numbers describes how the average of a randomly selected sample from a large population is likely to be close to t he average of the whole population. The second type of limit theorems that we st udy is known as central limit theorems. Central limit theorems are concerned wit h determining conditions under which the sum of a large number of random variabl es has a probability distribution that is approximately normal. 39 The Law of Large Numbers There are two versions of the law of large numbers: the weak law of large number s and the strong law of numbers. 39.1 The Weak Law of Large Numbers The law of large numbers is one of the fundamental theorems of statistics. One v ersion of this theorem, the weak law of large numbers, can be proven in a fairly straightforward manner using Chebyshevs inequality, which is, in turn, a special case of the Markov inequality. Our rst result is known as Markovs inequality. Proposition 39.1 (Markovs Inequality) If X 0 an c > 0, then P (X c) E(X) . c 397
398 Proof. Let c > 0. De ne I= 1 if X c 0 otherwise LIMIT THEOREMS Since X 0, I X . Taking expectations of both si e we lows since E(I) = P (X c) E(X) . c Example 39.1 From past experience a professor knows that the test score of a stu ent taking her nal examination is a ran om variable with mean 75. Give an upper boun for the probability that a stu ents test score will excee 85. Solution. Le t X be the ran om variable enoting the score on the nal exam. Using Markovs inequ ality, we have P (X 85) 75 E(X) = 0.882 85 85 Example 39.2 If X is a non negative ran om variable, then for all a > 0 we have P (X 1 aE(X)) a . Solution. The result follows by letting c = aE(X) is Markovs in equality Remark 39.1 Markovs inequality oes not apply for negative ran om variab le. To see this, let X be a ran om variable with range {1000, 1000}. Suppose that 1 P (X = 1000) = P (X = 1000) = 2 . Then E(X) = 0 an P (X 1000) = 0 Markovs boun gives us an upper boun on the probability that a ran om variable is large. It turns out, though, that there is a relate result to get an upper boun on the probability that a ran om variable is small. Proposition 39.2 Suppose that X is a ran om variable such that X M for some constant M. Then for all x < M we have P (X x) M E(X) M x n E(I) c Now the result fol

39 THE LAW OF LARGE NUMBERS Proof. By applying Markovs inequality we n P (X x) =P (M X M x) M E(X) E(M X) = M x M x 399 Example 39.3 Let X enote the test score of a ran om stu ent, where the maximum score obtainable is 100. Fin an upper boun of P (X 50), given that E(X) = 75. Solution. By the previous proposition we n P (X 50) 1 100 75 = 100 50 2 As a corollary of Proposition 39.1 we have Proposition 39.3 (Chebyshevs Inequalit y) If X is a ran om variable with nite mean an variance 2 , then for any value > 0, 2 P ( X ) 2 . Proof. Since (X )2 0, by Markovs inequality we can write But (X )2 to 2 2 ) E[(X )2 ] 2 . is equivalent to X an 2 = 2 2 Example 39.4 Show that for any random variable the probability of a deviation fr om the mean of more than k tandard deviation i le than or equal to k12 . So lution. Thi follow from Cheby hev inequality by u ing = k
this in turn is equivalent P ( X ) E[(X )2 ]

400 LIMIT THEOREMS Example 39.5 Suppo e X i the IQ of a random per on. We a ume X 0, E(X) = 100, and = 10. Find an upper bound of P (X 200) u ing r t Markov inequality and then C heby hev inequality. Solution. By u ing Markov inequality we nd 1 100 = 200 2 Now , u ing Cheby hev inequality we nd P (X 200) P (X 200) =P (X 100 100) =P (X 10) P ( X E(X) 10) 1 100
Example 39.6 On a ingle tank of ga oline, you expect to be able to drive 240 mi le before running out of ga . (a) Let p be the probability that you will NOT be able to make a 300 mile journey on a ingle tank of ga . What can you ay about p? (b) A ume now that you al o know that the tandard deviation of the di tanc e you can travel on a ingle tank of ga i 30 mile . What can you ay now about p? Solution. (a) Let X be the random variable repre enting mileage. Then by u i ng Markov inequality we nd p = P (X < 300) = 1 P (X 300) 1 (b) By Cheby hev in ality we nd p = P (X < 300) = 1P (X 300) 1P ( X240 60) 1 900 = 0.75 3600 240 00 Example 39.7 You toss a fair coin n times. Assume that all tosses are in epen en t. Let X enote the number of hea s obtaine in the n tosses. (a) Compute (expli citly) the variance of X. 9 (b) Show that P ( X E(X) n ) 4n . 3
39 THE LAW OF LARGE NUMBERS 401 Solution. (a) For 1 i n, let Xi = 1 if the ith toss shows hea s, an Xi = 0 othe rwise. Thus, X = X1 + X2 + + Xn . Moreover, E(Xi ) = 1 and E(Xi2 ) = 1 . Hence, 2 2 E(X) = n and 2 n 2 E(X ) =E i=1 2 Xi 2 = nE(X1 ) + i=j
E(Xi Xj ) = 1 n(n + 1) n + n(n 1) = 2 4 4 n(n+1) 4 Hence, Var(X) = E(X 2 ) (E(X))2 = (b) We apply Chebychevs inequality: P ( X E(X) n2 4 = n. 4 n Var(X) 9 ) = 2 3 (n/3) 4n When oes a ran om variable, X, have zero variance? It turns out that this happe ns when the ran om variable never eviates from the mean. The following theorem characterizes the structure of a ran om variable whose variance is zero. Proposi tion 39.4 If X is a ran om variable with zero variance, then X must be constant with probability equals to 1. Proof. First we show that if X 0 an E(X) = 0 then X = 0 an P (X = 0) = 1. Since E(X) = 0, by Markovs inequality P (X c) = 0 for a ll c > 0. But P (X > 0) = P 1 (X > ) n n=1 n=1 P (X > 1 ) = 0. n Hence, P (X > 0) = 0. Since X 0, 1 = P (X 0) = P (X = 0) + P (X > 0) = P (X = 0) . Now, suppose that V ar(X) = 0. Since (X E(X))2 0 an V ar(X) = E((X E(X))2 ), by the above result we have P (X E(X) = 0) = 1. That is, P (X = E(X)) = 1 One of the most well known an useful results of probability theory is the following t heorem, known as the weak law of large numbers.

402 LIMIT THEOREMS Theorem 39.1 Let X1 , X2 , , Xn be a sequence of indepedent random variables wit h common mean and nite common variance 2 . Then for any > 0 limn P or equivalently n X1 +X2 ++Xn n =0 lim P X1 + X2 + + X n < n =1 Proof. Since E X1 +X2 ++Xn n = and Var X1 +X2 ++Xn n = 2 n by Cheby hev inequality we nd 0P X 1 + X2 + + Xn n 2 n2
and the re ult follow by letting n ++Xn The above theorem says that for large n, X1 +X2n is small with high probabilit y. Also, it says that the distribution of the sample average becomes concentrate d near as n . Let A be an event with probability p. Repeat the experiment n times . Let ++Xn is the Xi be 1 if the event occurs and 0 otherwise. Then Sn = X1 +X2n nu mber of occurrence of A in n trials and = E(Xi ) = p. By the weak law of large n umbers we have n lim P ( Sn < ) = 1 The above statement says that, in a large number of repetitions of a Bernoulli e xperiment, we can expect the proportion of times the event will occur to be near p = P (A). This agrees with the de nition of probability that we introduced in Se ction 6 p. 54. The Weak Law of Large Numbers was given for a sequence of pairwis e independent random variables with the same mean and variance. We can generaliz e the Law to sequences of pairwise independent random variables, possibly with d i erent means and variances, as long as their variances are bounded by some consta nt.
39 THE LAW OF LARGE NUMBERS 403 Example 39.8 Let X1 , X2 , be pairwise independent random variables such that Va r(Xi ) b for some constant b > 0 and for all 1 i n. Let Sn = and n = E(Sn ). Show that, for every > 0 we have P ( Sn n > ) and consequently n X1 + X 2 + + X n n b 2 1 n lim P ( Sn n ) = 1 nd 0P Now,
Solution. Since E(Sn ) = n and Var(Sn ) = Chebyshevs inequality we Var(X1 )+Var(X2 )++Var(Xn ) n2 b . n bn n2 = b , n by X1 + X 2 + + Xn n n 1 2 1 P ( Sn n n b 2 1 n lim P ( Sn n ) = 1 ) = 1 P ( Sn n
> ) 1 By letting n we conclu e that
39.2 The Strong Law of Large Numbers Recall the weak law of large numbers: n lim P ( Sn < ) = 1
where the Xi s are independent identically distributed random variables and ++Xn Sn = X1 +X2n . This type of convergence is referred to as convergence
404 LIMIT THEOREMS
in probability. Unfortunately, this form of convergence does not assure converge nce of individual realizations. In other words, for any given elementary event x S, we have no assurance that limn Sn (x) = . Fortunately, however, there is a stro nger version of the law of large numbers that does assure convergence for indivi dual realizations. Theorem 39.2 Let {Xn }n1 be a sequence of in epen ent ran om v ariables with nite mean = E(Xi ) an K = E(Xi4 ) < . Then P X1 + X2 + + Xn = n = 1. Proof. We rst consider the case = 0 and let Tn = X1 + X2 + + Xn . Then 4 E(Tn ) = E[(X1 + X2 + + Xn )(X1 + X2 + + Xn )(X1 + X2 + + Xn )(X1 + X2 + + panding the product on the right side using the multinomial theorem the resultin g expression contains terms of the form Xi4 , Xi3 Xj , 2 Xi2 Xj , Xi2 Xj Xk and Xi Xj Xk Xl with i = j = k = l. Now recalling that = 0 and using the fact that the random va riables are independent we nd E(Xi3 Xj ) =E(Xi3 )E(Xj ) = 0 E(Xi2 Xj Xk ) =E(Xi2 )E(Xj )E(Xk ) = 0 E(Xi Xj Xk Xl ) =E(Xi )E(Xj )E(Xk )E(Xl ) = 0 Next, there are n terms of the form Xi4 and for each i = j the coe cient of 2 Xi2 Xj according to the multinomial theorem is 4! = 6. 2!2! But there are C(n, 2) = n(n1) di erent pair s of indices i = j. Thus, by taking 2 the expectation term by term of the expans ion we obtain 4 2 E(Tn ) = nE(Xi4 ) + 3n(n 1)E(Xi2 )E(Xj )
39 THE LAW OF LARGE NUMBERS 405 where in the last equality we made use of the independence assumption. Now, from the de nition of the variance we nd 0 V ar(Xi2 ) = E(Xi4 ) (E(Xi2 ))2 and this imp lies that (E(Xi2 ))2 E(Xi4 ) = K. It follows that 4 E(Tn ) nK + 3n(n 1)K which implies that E Therefore. 4 Tn 3K 4K K 3+ 2 2 4 n n n n E n=1 4 Tn n4 4K n=1 1 < n2 (39.1) Now, P n=1 4 Tn < +P n4 n=1 4 Tn = = 1. n4 If P able n=1 4 Tn n=1 n4 4 Tn n4 = > 0 then at some value in the range of the random varithe sum 4 Tn n=1 n4 is in nite and so its expected value is in nite 4 Tn n=1 n4 which contradicts (39.1). Hence, P = = 0 and therefore P n=1
4 Tn < = 1. n4 But the convergence of a series implies that its nth term goes to 0. Hence, P Si nce 4 Tn n4 4 Tn =0 =1 n n4 lim = Tn 4 n 4 = Sn the last result implies that P n lim Sn = 0 = 1
406 LIMIT THEOREMS which proves the result for = 0. Now, if = 0, we can apply the preceding argumen t to the random variables Xi to obtain n P or equivalently n lim i=1 (Xi ) =0 =1 n n P which proves the theorem n lim i=1 Xi = =1 n As an application of this theorem, suppose that a sequence of independent trials of some experiment is performed. Let E be a xed event of the experiment and let P (E) denote the probability that E occurs on any particular trial. De ne 1 if E o ccurs on the ith trial Xi = 0 otherwise By the Strong Law of Large Numbers we ha ve P X 1 + X2 + + Xn = E(X) = P (E) = 1 n n lim Since X1 +X2 + +Xn represents the number of times the event E occurs in the rst n t rials, the above result says that the limiting proportion of times that the even t occurs is just P (E). This justi es our de nition of P (E) that we introduced in S ection 6,i.e., P (E) = lim n(E) n n where n(E) denotes the number of times in the rst n repetitions of the experiment that the event E occurs. To clarify the somewhat subtle di erence between the Wea k and Strong Laws of Large Numbers, we will construct an example of a sequence X 1 , X2 , of mutually independent random variables that satis es the Weak Law of La rge but not the Strong Law.
39 THE LAW OF LARGE NUMBERS 407 Example 39.9 Let X1 , X2 , be the sequence of mutually independent random variab les such that X1 = 0, and for each positive integer i P (Xi = i) = 1 , 2i ln i P (Xi = i) = 1 , 2i ln i P (Xi = 0) = 1 1 i ln i Tn . n Note that E(Xi ) = 0 for all i. Let Tn = X1 + X2 + + Xn and Sn = 2 n (a) Show that Var(Tn ) ln n . (b) Show that the sequence X1 , X2 , satis es the Weak Law of Large Numbers, i.e., prove that for any > 0 n lim P ( Sn ) = 0.
(c) Let A1 , A2 , be any in nite sequence of mutually independent events such that P (Ai ) = . i=1 Prove that P(in nitely many Ai occur) = 1 (d) Show that P ( Xi clu e that lim P (Sn = ) = 0 n i) = . i=1 (e) Con
an hence that the Strong Law of Large Numbers completely fails for the sequence X1 , X2 , Solution. (a) We have Var(Tn ) =Var(X1 ) + Var(X2 ) + + Var(Xn ) n =0 + i=2 n (E(Xi2 ) (E(Xi ))2 ) i2 1 i ln i = i=2 n = i=2 i ln i
408 LIMIT THEOREMS x 1 1 Now, if we let f (x) = ln x then f (x) = ln x 1 ln x > 0 for x > e so i n that f (x) is increasing for x > e. It follows that ln n ln i for 2 i n. Further more, n2 = ln n Hence, n i=1 n ln n n i=2 i ln i Var(Tn ) (b) We have n2 . ln n
P ( Sn ) =P ( Sn 0 ) 1 Var(Sn ) 2 Chebyshev sinequality Var(Tn ) 1 2 n2 1 2 n = which goes to zero as n goes to in nity. (c) Let Tr,n = n IAi the number of e vents Ai with r i n that occur. i=r Then n n n lim E(Tr,n ) = lim n E(IAi ) = lim i=r n P (Ai ) = i=r
Since ex 0 as x we conclude that eE(Tr,n ) 0 as n . Now, let Kr be the event t o Ai with i r occurs. Also, let Kr,n be the event that no Ai with r i n occurs. Finally, let K be the event that only nitely many Ai s. occurs. We must prove tha t P (K) = 0. We rst show that P (Kr,n ) eE(Tr,n ) .
39 THE LAW OF LARGE NUMBERS We remin the rea er of the inequality 1 + x ex . We have P (Kr,n ) =P (Tr,n = 0) = P [(Ar Ar+1 An )c ] =P [Ac Ac Ac ] n 409 = i=r n P (Ac ) i [1 P (Ai )] i=r n = =e i=r eP (Ai ) n i=r n i=r P (Ai ) E(IAi ) =e =eE(Tr,n ) Now, since Kr Kr,n we conclude that 0 P (Kr ) P (Kr,n ) eE(Tr,n ) 0 n . Hence, P (Kr ) = 0 for all r n. Now note that K = r Kr so by Booles inequality (See Proposition 35.3), 0 P (K) r Kr = 0. That is, P (K) = 0. Hence the probabil ity that in nitely many Ai s occurs is 1. 1 (d) We have that P ( Xi i) = i ln i . Thus, n n P ( Xi i) =0 + i=1 n i=2 1 i ln i
x 2 x ln x = ln ln n ln ln 2
an this last term approaches in nity as n approaches in nity. (e) By parts (c) an ( ), the probability that Xi i for in nitely many i is 1. But if Xi i for in n itely many i then from the e nition of limit limn Xn = 0. Hence, n Xn P ( lim = 0) = 1 n n which means Xn P ( lim = 0) = 0 n n
410 Now note that LIMIT THEOREMS
Xn n1 = Sn Sn1 n n so that if limn Sn = 0 then limn Xn = 0. This implies that n P ( m Sn = 0) P ( lim n Xn = 0) n n That is, P ( lim Sn = 0) = 0 n an this violates the Stong Law of Large numbers
39 THE LAW OF LARGE NUMBERS 411 Practice Problems Problem 39.1 Let > 0. Let X be a iscrete ran om variable with range { , } an pm f given by p( ) = 1 an p( ) = 1 an 0 otherwise. Show that the inequality 2 2 in Chebyshevs inequality is in fact an equality. Problem 39.2 Let X1 , X2 , , Xn be a Bernoulli trials process with probability 0.3 for success and 0.7 for failure . Let Xi = 1 if the ith outcome is a success and 0 otherwise. Find n so that P S n E Sn 0.1 0.21, where n n S n = X 1 + X2 + + Xn . Problem 39.3 Suppose that X s uniformly distributed on [0, 12]. Find an upper bound for the probability that a sample from X lands more than 1.5 standard deviation from the mean. Problem 3 9.4 Consider the transmission of several 10Mbytes les over a noisy channel. Suppo se that the average number of erroneous bits per transmitted le at the receiver o utput is 103 . What can be said about the probability of having 104 erroneous bi ts uring the transmission of one of these les? Problem 39.5 Suppose that X is a ran om variable with mean an variance both equal to 20. What can be sai about P (0 < X < 40)? Problem 39.6 Let X1 , X2 , , X20 be independent Poisson random v ariables with mean 1. Use Markovs inequality to obtain a bound on P (X1 + X2 + + X20 > 15). Problem 39.7 From past experience a professor knows that the test sco re of a student taking her nal examination is a random variable with mean 75 and variance 25. What can be said about the probability that a student will score be tween 65 and 85?

412 LIMIT THEOREMS Problem 39.8 Let MX (t) = E(etX ) be the moment generating function of a random variable X. Show that P (X ) e t MX (t). Problem 39.9 On average a shop sells 50 telephones each week. Stock one Mon ay is 75. What can be sai about the probabi lity of not having enough stock left by the en of the week? Problem 39.10 Let X be a ran om variable with mean you say about P (0.475 X 0.525)? 1 2 an variance 25 107 . What can Problem 39.11 1 Let X 0 be a ran om variable with mean . Show that P (X 2) 2 . Pro blem 39.12 The numbers of robots pro uce in a factory uring a ay is a ran om variable with mean 100. Describe the probability that the factorys pro uction wil l be more than 120 in a ay. Also if the variance is known to be 5, then escrib e the probability that the factorys pro uction will be between 70 an 130 in a a y. Problem 39.13 A biase coin comes up hea s 30% of the time. The coin is tosse 400 times. Let X be the number of hea s in the 400 tossings. Use Chebyshevs ine quality to boun the probability that X is between 100 an 140. Problem 39.14 Le t X1 , , Xn be independent random variables, each with probability density funct ion: 2x 0 x 1 f (x) = 0 elsewhere n Show that X = i=1 n Xi converges in probability to a constant as n and nd that constant.
39 THE LAW OF LARGE NUMBERS 413 Problem 39.15 Let X1 , , Xn be independent and identically distributed Uniform(0 ,1). Let Yn be the minimum of X1 , , Xn . (a) Find the cumulative distribution o f Yn (b) Show that Yn converges in probability to 0 by showing that for arbitrar y >0 lim P ( Yn 0 ) = 1. n
414 LIMIT THEOREMS
40 The Central Limit Theorem The central limit theorem is one of the most remarkable theorem among the limit theorems. This theorem says that the sum of large number of independent identica lly distributed random variables is well approximated by a normal random variabl e. We rst need a technical result. Theorem 40.1 Let Z1 , Z2 , be a sequence of ra ndom variables having distribution functions FZn and moment generating functions MZn , n 1. Let Z be a ran om variable having istribution FZ an moment generat ing function MZ . If MZn (t) MZ (t) as n an for all t, then FZn FZ for all t at which FZ (t) is continuous. With the above theorem, we can prove the central li mit theorem. Theorem 40.2 Let X1 , X2 , be a sequence of independent and identic ally distributed random variables, each with mean and variance 2 . Then, a x2 1 n X1 + X 2 + + X n a e 2 dx P n 2 as n . The Central Limit Theorem s of the underlying distribution of the variables Xi , so long as they are inde endent, the distribution of n X1 +X2 ++Xn converges to the same, normal, distribut on. n Proof. We prove the theorem under the a umption that E(etXi ) i nite in a neighborhood of 0. In particular, we how that M n ( X1 +X2 ++Xn ) (x) e 2 n x2 x2
where e 2 i
the moment generating function of the tandard normal di tribution.

40 THE CENTRAL LIMIT THEOREM Now, u ing the propertie of moment generating func tion we can write M n ( X1 +X2 ++Xn ) (x) =M 1 n n Xi n i=1 415 (x) =M n Xi n i=1 x n x n x n n n = i=1 M Xi = M X1 = MY where Y = x n
X1 . x n = MY (0) + MY (0) x n + 1 x2 M (0) + R 2n Y x n , 0 a x 0 But E(Y ) = E and E(Y ) = E 2 X1 =0 2 X1 =
Now expand MY (x/ n) in a Taylor n R x2 x n
erie around 0 a follow
MY where
V ar(X1 ) = 1. 2 By Propo ition 38.1 we obtain MY (0) = 1, MY (0) = E(Y ) = 0, and MY (0) = E(Y 2 ) = 1. Hence, MY x n =1+ 1 x2 +R 2n x n
416 and o lim MY x n n LIMIT THEOREMS n = lim = lim n n 1 x2 x 1+ +R 2n n 2 1 x x 1+ + nR n 2 n n R x2 x n n n n But nR a n . Hence, x n lim = x2 0 n MY x n =e2. X1 +X2 ++Xn n x2 Now the result follows from Theorem 40.1 with Zn = n Z tandard normal di tributi on, n X1 + X 2 + + Xn FZn (a) = P n and FZ (a) = (a) = a , a e 2 dx x2
The central limit theorem suggests approximating the random variable n X1 + X2 + + Xn n with a tandard normal random variable. Thi implie that the ample mea 2 ha approximately a normal di tribution with mean and variance . n Al o, a u m of n independent and identically di tributed random variable with common mean and variance 2 can be approximated by a normal di tribution with mean n and vari ance n 2 . Example 40.1 The weight of an arbitrary airline pa enger baggage ha a mean of 20 pound and a variance of 9 pound . Con ider an airplane that carrie 200 pa enger , and a ume every pa enger check one piece of luggage. E tima te the probability that the total baggage weight exceed 4050 pound .
40 THE CENTRAL LIMIT THEOREM Solution. Let Xi =weight of baggage of pa enger i. Thu , 200 417 P i=1 Xi > 4050 =P 200 i=1 Xi 200(20) 4050 20(200) > 3 200 3 200 P (Z > 1.179) = 1 P (Z 1.179) =1 (1.179) = 0.1314 Example 40.2 In an analysis of ealthca e data, ages have been ounded to the nea est multiple of 5 yea s. The d i e ence between the t ue age and the ounded age is assumed to be unifo mly dist ibuted on the inte val f om 2.5 yea s to 2.5 yea s. The healthca e data a e based on a andom sample of 48 people. What is the app oximate p obability that the m ean of the ounded ages is within 0.25 yea s of the mean of the t ue ages? Solut ion. Let X denote the di e ence between t ue and epo ted age. We a e given X is u nifo mly dist ibuted on (2.5, 2.5). That is, X has pdf f (x) = 1/5, 2.5 < x < 2.5. It follows that E(X) = 0 and 2.5 2 X = E(X 2 ) = 2.5 x2 dx 2.083 5 o that SD(X) = 2.083 1.443. Now X 48 the di erence between the mean of the true and rounded age , ha a di tribution that i approximately normal with mean 0 an d tandard deviation 1.443 0.2083. Therefore, 48 P 1 1 X 48 4 4 0.25 X48 0.25 083 0.2083 0.2083 =P (1.2 Z 1.2) = 2(1.2) 1 2(0.8849) 1 = 0.77 =P Example 40.3 A computer generates 48 random real numbers, rounds each number to the nearest integer and then computes the average o these 48 rounded values. As sume that the numbers generated are independent o each other and that the round ing errors are distributed uni ormly on the interval [0.5, 0.5]. ind the approxi mate probability that the average o the rounded values is within 0.05 o the av erage o the exact numbers.

418 LIMIT THEOREMS Solution. 1 Let X1 , , X48 denote the 48 rounding errors, and X = 48 48 Xi their i=1 average. We need to compute P ( X 0.05). Since a rounding error is uniform ly distributed on [0.5, 0.5], its mean is = 0 and its variance is 2 = 0.5 0.5 x2 dx = x3 3 0.5 0.5 = 1 . 12 2
P ( X 0.05) P (24 (0.05) 24X 24 (0.05)) =(1.2) (1.2) = 2(1.2) 1 = 0.76 Let X1 , X2 , X3 , X4 be a random sample o size 4 rom a normal distribution w ith mean 2 and variance 10, and let X be the sample mean. Determine a such that P (X a) = 0.90. Solution. The sample mean X is normal with mean = 2 and variance and standard deviation 2.5 1.58, so 0.90 = P (X a) = P rom the normal table, w e get X 2 a2 < 1.58 1.58 a2 1.58 2 n 10 4 = = 2.5, = a2 1.58 . = 1.28, so a = 4.02 Example 40.5 Assume that the weights o individuals are independent and normally distributed with a mean o 160 pounds and a standard deviation o 30 pounds. Su ppose that 25 people s ueeze into an elevator that is designed to hold 4300 poun ds. (a) What is the probability that the load (total weight) exceeds the design limit? (b) What design limit is exceeded by 25 occupants with probability 0.001? Solution. (a) Let X be an individuals weight. Then, X has a normal distribution with = 160 pounds and = 30 pound . Let Y = X1 + X2 + + X25 , where
1 approximate di tribution N (, ) = N (0, 242 ). Thu 24X i approximately n dard normal, o
tan
By the Central Limit Theorem, X ha
40 THE CENTRAL LIMIT THEOREM 419 Xi denotes the ith persons weight. Then, Y has a normal distribution with E(Y ) = 25E(X) = 25 (160) = 4000 pounds and Var(Y ) = 25Var(X) = 25 (900) = 22500. Now, the desired probability is P (Y > 4300) =P Y 4000 4300 4000 > 22500 22500 =P (Z > 2) = 1 P (Z 2) =1 0.9772 = 0.0228 (b) We want to nd x such that P (Y > x) = 0.001. Note that P (Y > x) =P =P Y 4000 x 4000 > 22500 22500 x 4000 Z> = 0.01 22500 It is equivalent to P Z x4000 = 0.999. From the normal Table we nd 22500 P (Z 3.09 ) = 0.999. So (x 4000)/150 = 3.09. Solving for x we nd x 4463.5 pounds
420 Practice Problems LIMIT THEOREMS Problem 40.1 A manufacturer of booklets packages the booklets in boxes of 100. I t is known that, on average, the booklets weigh 1 ounce, with a standard deviati on of 0.05 ounces. What is the probability that 1 box of booklets weighs more th an 100.4 ounces? Problem 40.2 In a particular basketball league, the standard de viation in the distribution of players height is 2 inches. Twenty ve players are s elected at random and their heights are measured. Give an estimate for the proba bility that the average height of the players in this sample of 25 is within 1 i nch of the league average height. Problem 40.3 A lightbulb manufacturer claims t hat the lifespan of its lightbulbs has a mean of 54 months and a st. deviation o f 6 months. Your consumer advocacy group tests 50 of them. Assuming the manufact urers claims are true, what is the probability that it nds a mean lifetime of less than 52 months? Problem 40.4 If 10 fair dice are rolled, nd the approximate prob ability that the sum obtained is between 30 and 40, inclusive. Problem 40.5 Let Xi , i = 1, 2, , 10 be independend random variables each uniformly distributed o ver (0,1). Calculate an approximation to P ( 10 Xi > 6). i=1 Problem 40.6 Suppos e that Xi , i = 1, , 100 are exponentially distributed random vari100 X 1 ables with parameter = 10000 . Let X = i=1 i . Approximate P (950 100 X 1050). Prob em 40.7 The Braves p ay 100 independent baseba games, each of which they have pr obabi ity 0.8 of winning. Whats the probabi ity that they win at east 90?
40 THE CENTRAL LIMIT THEOREM 421 Prob em 40.8 An insurance company has 10,000 automobi e po icyho ders. The expec ted year y c aim per po icyho der is $240 with a standard deviation of $800. App roximate the probabi ity that the tota year y c aim exceeds $2.7 mi ion. Prob em 40.9 A certain component is critica to the operation of an e ectrica system and must be rep aced immediate y upon fai ure. If the mean ife of this type of component is 100 hours and its standard deviation is 30 hours, how many of thes e components must be in stock so that the probabi ity that the system is in cont inua operation for the next 2000 hours is at east 0.95? Prob em 40.10 Student scores on exams given by a certain instructor has mean 74 and standard deviation 14. This instructor is about to give two exams. One to a c ass of size 25 and t he other to a c ass of 64. Approximate the probabi ity that the average test sco re in the c ass of size 25 exceeds 80. Prob em 40.11 A cha ity eceives 2025 con t ibutions. Cont ibutions a e assumed to be independent and identically dist ibu ted with mean 3125 and standa d deviation 250. Calculate the app oximate 90th pe centile fo the dist ibution of the total cont ibutions eceived. P oblem 40.12 An insu ance company issues 1250 vision ca e insu ance policies. The numbe of claims led by a policyholde unde a vision ca e insu ance policy du ing one yea is a Poisson andom va iable with mean 2. Assume the numbe s of claims led by di stinct policyholde s a e independent of one anothe . What is the app oximate p o bability that the e is a total of between 2450 and 2600 claims du ing a one yea pe iod? P oblem 40.13 A company manufactu es a b and of light bulb with a lifet ime in months that is no mally dist ibuted with mean 3 and va iance 1 . A consum e buys a numbe of these bulbs with the intention of eplacing them successivel y as they bu n out. The light bulbs have independent lifetimes. What is the smal lest numbe of bulbs to be pu chased so that the succession

422 LIMIT THEOREMS of light bulbs p oduces light fo at least 40 months with p obability at least 0 .9772? P oblem 40.14 Let X and Y be the numbe of hou s that a andomly selected pe son watches movies and spo ting events, espectively, du ing a th ee month p e iod. The following info mation is known about X and Y : E(X) = 50 E(Y) = 20 Va (X) = 50 Va (Y) = 30 Cov (X,Y) = 10 One hund ed people a e andomly selected an d obse ved fo these th ee months. Let T be the total numbe of hou s that these one hund ed people watch movies o spo ting events du ing this th ee month pe i od. App oximate the value of P (T < 7100). P oblem 40.15 The total claim amount fo a health insu ance policy follows a dist ibution with density function f (x) = x 1 e 1000 1000 0 x>0 othe wise The p emium fo the policy is set at 100 ove the expected total claim amount. I f 100 policies a e sold, what is the app oximate p obability that the insu ance company will have claims exceeding the p emiums collected? P oblem 40.16 A city has just added 100 new female ec uits to its police fo ce. The city will p ovid e a pension to each new hi e who emains with the fo ce until eti ement. In add ition, if the new hi e is ma ied at the time of he eti ement, a second pensio n will be p ovided fo he husband. A consulting actua y makes the following ass umptions: (i) Each new ec uit has a 0.4 p obability of emaining with the polic e fo ce until eti ement. (ii) Given that a new ec uit eaches eti ement with the police fo ce, the

40 THE CENTRAL LIMIT THEOREM 423 p obability that she is not ma ied at the time of eti ement is 0.25. (iii) The numbe of pensions that the city will p ovide on behalf of each new hi e is ind ependent of the numbe of pensions it will p ovide on behalf of any othe new hi e. Dete mine the p obability that the city will p ovide at most 90 pensions to the 100 new hi es and thei husbands. P oblem 40.17 Delive y times fo shipments f om a cent al wa ehouse a e exponentially dist ibuted with a mean of 2.4 days (note that times a e measu ed continuously, not just in numbe of days). A ando m sample of n = 100 shipments a e selected, and thei shipping times a e obse ve d. What is the app oximate p obability that the ave age shipping time is less th an 2.0 days? P oblem 40.18 (a) Give the app oximate sampling dist ibution fo th e following quantity based on andom samples of independent obse vations: X= Xi , E(Xi ) = 100, Va (Xi ) = 400. 100 100 i=1 (b) What is the app oximate p obability the sample mean will be between 96 and 1 04? P oblem 40.19 A biased coin comes up heads 30% of the time. The coin is toss ed 400 times. Let X be the numbe of heads in the 400 tossings. (a) Use Chebyshe vs inequality to bound the p obability that X is between 100 and 140. (b) Use no mal app oximation to compute the p obability that X is between 100 and 140.

424 LIMIT THEOREMS
41 Mo e Useful P obabilistic Inequalities The impo tance of the Ma kovs and Chebyshevs inequalities is that they enable us t o de ive bounds on p obabilities when only the mean, o both the mean and the va iance, of the p obability dist ibution a e known. In this section, we establish mo e p obability bounds. The following esult gives a tighte bound in Chebyshe vs inequality. P oposition 41.1 Let X be a andom va iable with mean and nite va i ance 2 . Then for any a > 0 2 P (X + a) 2 + a2 and 2 P (X a) 2 + a2 P lo of generality we a ume that = 0. Then for any b > 0 we have P (X a) =P (X + b a + b) P ((X + b)2 (a + b)2 ) E[(X + b)2 ] (a + b)2 2 + b2 = (a + b)2 + t2 = g(t) (1 + t)2 where = Since g (t) = 2 a2 b and t = a . 1 t2 + (1 )t 2 (1 + t)4 we nd g (t) = 0 when t = . Since g (t) = 2(2t + 1 )(1 + t)4 8(t2 + (1 )t )(1 g () = 2( + 1)3 > 0 so th t t = is the

41 MORE USEFUL PROBABILISTIC INEQUALITIES minimum of g(t) with g() = It follows t h t 2 = 2 . 1+ + a2 425 2 . 2 + a2 Now, uppo e that = 0. Since E(X E(X)) = 0 and V ar(X E(X)) = V ar(X) = 2 , by applying the previou inequality to X we obtain P (X a) P (X + a) + a2
Similarly, ince E( X) = 0 and V ar( X) = V ar(X) = 2 , we get P ( X a) or P 2 + a2 2 2 + a2 Example 41.1 If the number produced in a factory during a week i a random varia ble with mean 100 and variance 400, compute an upper bound on the probability th at thi week production will be at lea t 120. Solution. Applying the previou re ult we nd P (X 120) = P (X 100 20) 400 1 = 400 + 202 2 The following provide bound on P (X a) in term of the moment generating funct ion M (t) = etX with t > 0. Propo ition 41.2 (Cherno bound) Let X be a random va riable and uppo e that M (t) = E(etX ) i nite. Then P (X a) eta M (t), t > 0 and P (X a) eta M (t), t<0
LIMIT THEOREMS
where the la t inequality follow from Markov inequality. Similarly, for t < 0 w e have P (X a) P (etX eta ) E[etX ]eta It follow from Cherno inequality that a p bound for P (X a) i a minimizer of the function eta M (t). Example 41.2 Let Z i a tandard random variable o that it moment generating function t2 i M (t) = e 2 . Find a harp upper bound for P (Z a). Solution. By Cherno inequality we h ave P (Z a) eta e 2 = e 2 ta , t > 0 Let g(t) = e 2 ta . Then g (t) = (t a)e 2 ta that g (t) = 0 when t = a. Since g (t) = e 2 ta + (t a)2 e 2 ta we nd g (a) > 0 o that t = a i the minimum of g(t). Hence, a harp bound i P (Z a) e 2 , t > 0 S imilarly, for a < 0 we nd P (Z a) e 2 , t < 0 The next inequality i one having to do with expectation rather than probabilitie . Before tating it, we need the following de nition: A di erentiable function f (x) i aid to be convex on the open interval I = (a, b) if f (u + (1 )v) f (u) + (1 )f (v) for ll u nd v in I nd . Geometric lly, this s ys th t the gr ph of f (x) lies completely bove e ch t ngent line. 2 2 t2 t2 t2 t2 t2 t2
426 Proof. Suppo e
r t that t > 0. Then P (X a) P (etX eta ) E[etX ]eta
41 MORE USEFUL PROBABILISTIC INEQUALITIES Proposition 41.3 (Jensens inequ lity) I f f (x) is convex function then E(f (X)) f (E(X)) provi e that the expectatio ns exist an are nite. Proof. The tangent line at (E(x), f (E(X))) is y = f (E(X) ) + f (E(X))(x E(X)). By convexity we have f (x) f (E(X)) + f (E(X))(x E(X)). Up on taking expectation of both si es we n E(f (X)) E[f (E(X)) + f (E(X))(X E(X))] =f (E(X)) + f (E(X))E(X) f (E(X))E(X) = f (E(X)) Example 41.3 Let X be a ran om variable. Show that E(eX ) eE(X) . 427 Solution. Since f (x) = ex is convex, by Jensens inequality we can write E(eX ) e E(X) Example 41.4 Suppose that {x1 , x2 , , xn } is a set of positive numbers. S how that the arithmetic mean is at least as large as the geometric mean: (x1 x2 xn ) n 1 1 (x1 + x2 + + xn ). n 1 n Solution. Let X be a random variable such that p(X = xi ) = g(x) = ln x. By Jens ens inequality we have E[ ln X] ln [E(X)]. for 1 i n. Let

428 That is E[ln X] ln [E(X)]. But E[ln X] = an n LIMIT THEOREMS ln xi = i=1 1 ln (x1 x2 xn ) n
1 ln [E(X)] = ln (x1 + x2 + + xn ). n 1 1 ln (x1 x2 xn ) n ln (x1 + x2 + + xn ) n It follows that Now the result follows by taking ex of both sides and recalling that ex is an in creasing function
1 n
41 MORE USEFUL PROBABILISTIC INEQUALITIES 429 Practice Problems Problem 41.1 Roll a single fair die and let X be the outcome. Then, E(X) = 3.5 a nd V ar(X) = 35 . 12 (a) Compute the exact value of P (X 6). (b) Use Markovs ineq uality to n an upper boun of P (X 6). (c) Use Chebyshevs inequality to n an uppe r boun of P (X 6). ( ) Use one si e Chebyshevs inequality to n an upper boun o f P (X 6). Problem 41.2 Fin Cherno boun s for a binomial ran om variable with pa rameters (n, p). Problem 41.3 Suppose that the average number of parking tickets given in the Engineering lot at Stony Brook is three per ay. Also, assume that you are tol that the variance of the number of tickets in any one ay is 9. Gi ve an estimate of the probability p, that at least ve parking tickets will be giv en out in the Engineering lot tomorrow. Problem 41.4 Each time I write a check, I omit the cents when I recor the amount in my checkbook (I recor only the int eger number of ollars). Assume that I write 20 checks this month. Fin an upper boun on the probability that the recor in my checkbook shows at least $15 les s than the actual amount in my account. Problem 41.5 Fin the cherno boun s for a Poisson ran om variable X with parameter . Prob em 41.6 Let X be a Poisson rando m variab e with mean 20. (a) Use the Markovs inequa ity to obtain an upper bound on p = P (X 26). (b) Use the Cherno bound to obtain an upper bound on p. (c) Use the Chebyshevs bound to obtain an upper bound on p. (d) Approximate p by making u se of the centra imit theorem.

430 Prob em 41.7 Let X be a random variab e. Show the fo owing (a) E(X 2 ) [E(X )]2 . 1 1 (b) If X 0 then E X E(X) . (c) If X > 0 then E[ n X] n [E(X)] LIMIT THEOREMS Prob em 41.8 a Let X be a random variab e with density function f (x) = xa+1 , x 1, a > 1. We ca X a pareto random variab e with parameter a. (a) Find E(X). 1 (b) Find E X . 1 (c) Show that g(x) = x is convex in (0, ). (d) Verify Jensens in equa ity by comparing (b) and the reciproca of (a). Prob em 41.9 Suppose that { x1 , x2 , , xn } is a set of positive numbers. Prove (x1 x2 xn ) n 2 x2 + x2 + + x2 1 2 n . n
Appendix 42 Improper Integrals A very common mistake among students is when evaluating the integral 1 1 dx. A n on careful student will just argue as follows 1 x 1 1 1 dx = [ln x ]1 = 0. 1 x Unfortunately, thats not the right answer as we will see below. The important fac t ignored here is that the integrand is not continuous at x = 0. b Recall that t he de nite integral a f (x)dx was de ned as the limit of a leftor right Riemann sum. We noted that the de nite integral is always wellde ned if: (a) f (x) is continuous on [a, b], (b) and if the domain of integration [a, b] is nite. Improper integra ls are integrals in which one or both of these conditions are not met, i.e., (1) The interval of integration is in nite: [a, ), (, b], (, ), e.g.: 1 1 dx. x (2) The integrand has an in nite discontinuity at some point c in the interval [a, b], i.e. the integrand is unbounded near c : xc lim f (x) = . 431
432 e.g.: APPENDIX 1 dx. 0 x An improper integral may not be well de ned or may have in nite value. In this case we say that the integral is divergent. In case an improper integral ha s a nite value then we say that it is convergent. We will consider only improper integrals with positive integrands since they are the most common. Unbounded Int ervals of Integration The rst type of improper integrals arises when the domain o f integration is in nite. In case one of the limits of integration is in nite, we de n e b 1 f (x)dx = lim a b f (x)dx a b or b f (x)dx = lim a f (x)dx. a If both limits are in nite, then we choose any number c in the domain of f and de ne c f (x)dx = f (x)dx + f (x)dx. c In this case, the integral is convergent if and only if both integrals on the ri ght converge. Alternatively, we can also write R f (x)dx = lim R f (x)dx. R Example 42.1 Does the integral Solution. We have 1 1 dx 1 x2 converge or diverge? 1 dx = lim b x2
b 1 1 1 1 dx = lim [ ]b = lim ( + 1) = 1. 1 2 b b x x b In terms of area, the given integral represents the area under the graph of 1 f (x) = x2 from x = 1 and extending in nitely to the right. The above improper integ ral says the following. Let b > 1 and obtain the area shown in
42 IMPROPER INTEGRALS Figure 42.1. 433 Figure 42.1 1 Then 1 x2 dx is the area under the graph of f (x) from x = 1 to x = b. As b ge ts larger and larger this area is close to 1. b Example 42.2 Does the improper integral Solution. We have 1 1 dx 1 x converge or diverge? 1 dx = lim b x b 1 1 dx = lim [2 x]b = lim (2 b 2) = . 1 b b x So the improper integral is divergent. Remark 42.1 In general, some unbounded re gions have nite areas while others have in nite areas. This is true whether a regio n extends to in nity along the x axis or along the y axis or both, and whether it extends to in nity on one or both 1 sides of an axis. For example the area under t he graph of x2 from 1 to in nity 1 is nite whereas that under the graph of x is in nit e. This has to do with how fast each function approaches 0 as x . The function f (x) = x2 grows very rapidly which means that the graph is steep. When we conside r 1 the reciprocal x2 , it thus approaches the x axis very quickly and so the ar ea bounded by this graph is nite. 1 On the other hand, the function f (x) = x 2 g rows very slowly meaning that its graph is relatively at. This means that the gra ph y = 11 approaches the x2 x axis very slowly and the area bounded by the graph is in nite.
434 APPENDIX The following example generalizes the results of the previous two examples. Exam ple 42.3 Determine for which values of p the improper integral Solution. Suppose rst that p = 1. Then 1 dx 1 x 1 = limb 1 x dx = limb [ln x ]b = limb ln b = 1 b 1 dx 1 xp diverges. so the improper integral is divergent. Now, suppose that p = 1. Then 1 dx 1 xp = limb 1 xp dx p+1 = limb [ x ]b p+1 1 p+1 1 = limb ( b p+1 ). p+1 b If p < 1 then p + 1 > 0 so that limb bp+1 = and therefore the improper integral is d ivergent. If p > 1 then p+1 < 0 so that limb bp+1 = 0 and hence the improper integra l converges: 1 1 1 dx = . p x p + 1 cx e dx 0 Example 42.4 For what values of c is the improper integral Solution. We have convergent? cx e dx 0 = limb 0 ecx dx = limb 1 ecx b 0 c = limb 1 (ecb 1) = 1 c c b provided that c < 0. Otherwise, i.e. if c 0, then the improper integral is iver gent. Remark 42.2 The previous two results are very useful an you may want to m emorize them.
42 IMPROPER INTEGRALS Example 42.5 Show that the improper integral 435 1 x 1+x2 converges. Solution. Splitting the integral into two as follows: 1 x = 1 + x2 0 1 x + 1 + x2 0 1 x. 1 + x2
Now, 0 1 x 1+x2 1 = lima a 1+x2 x = lima arctan x 0 a = lima (arctan 0 arctan a) 1 dx 0 1+x2 0 Similarly, we nd that = 2 so that 1 dx 1+x2 = 2 + 2 = . Unbounded Integrands Su Then we de ne b b a f (x)dx = lim + ta f (x)dx. t Similarly, if f (x) is unbounded at x = b, that is limxb f (x) = . Then we de ne b t a f (x)dx = lim tb
ose f (x) is unbounded at x = a, that is limxa+ f (x) = .

f (x)dx. a Now, if f (x) is unbounded at an interior oint a < c < b then we de ne b a t b f (x)dx = lim tc a f (x)dx + lim + tc f (x)dx. t If both limits exist then the integral on the left hand side converges. If one o f the limits does not exist then we say that the im ro er integral is divergent. Exam le 42.6 Show that the im ro er integral Solution. Indeed, 1 1 dx 0 x 1 1 dx 0 x converges. 1 1 = limt0+ t x dx = limt0+ 2 x 1 t = limt0+ (2 2 t) = 2.
436 In terms of area, we pick an t > 0 as shown in Figure 42.2: APPENDIX Figure 42.2 Then the shaded area is approaches the value 2. 1 1 dx. t x As t approaches 0 from the right, the area Example 42.7 Investigate the convergence of 2 1 dx. 0 (x2)2
Solution. We deal with this improper integral as follows 2 1 dx 0 (x2)2 1 1 = limt2 0 (x2)2 dx = limt2 (x2) t 0 1 = limt2 ( t2 1 ) = So that the given improper integral is divergent. Example 42.8 Investigate the i mproper integral Solution. We rst write 1 1 dx. 1 x 1 1 1 dx = x 0 1 1 dx + x 1 0 1 dx. x On one hand we have, 0 1 dx 1 x 1 = limt0 1 x dx = limt0 ln x t 1 = limt0 ln t = . 0 1 dx 1 x t This shows that the improper integral 1 1 improper integral 1 x dx is divergent. is divergent and therefore the
42 IMPROPER INTEGRALS 437 Improper Integrals of Mixed Types Now, if the interval of integration is unbound ed and the integrand is unbounded at one or more points inside the interval of i ntegration we can evaluate the improper integral by decomposing it into a sum of an improper integral with nite interval but where the integrand is unbounded and an improper integral with an in nite interval. If each component integrals conver ges, then we say that the original integral converges to the sum of the values o f the component integrals. If one of the component integrals diverges, we say th at the entire integral diverges. Example 42.9 Investigate the convergence of 1 dx. 0 x2 Solution. Note that the interval of integration is in nite and the function is und e ned at x = 0. So we write 0 1 dx = x2 1 t 1 0 1 dx + x2 1 1 dx. x2 But 0 1 1 dx = lim t0+ x2 1 1 1 dx = lim 1 = lim ( 1) = . t 2 + + t t0 t0 x x 1 dx 0 x2 Thus, diverges. 1 1 dx 0 x2 diverges and consequently the improper integral
438 APPENDIX 43 Double Integrals In this section, we introduce the concept of de nite integral of a function of two variables over a rectangular region. By a rectangular region we mean a region R as shown in Figure 43.1(I).
Figure 43.1 Let f (x, y) be a continuous function on R. Our de nition of the de nite integral of f over the rectangle R will follow the de nition from one variable ca lculus. Partition the interval a x b into n equal subintervals using the mesh po ints a x0 < x1 < x2 < < xn1 < xn = b with x = ba n denoting the length of each s nterval. Similarly, partition c y d into m subintervals using the mesh points c = y0 < y1 < y2 < < ym1 < ym = d with y = dc denoting the length of each subinterva . This way, m the rectangle R is partitioned into mn subrectangles each of area equals to xy as shown in Figure 43.1(II). Let Dij be a typical rectangle. Let mij be the smallest value of f in Dij and Mij be the largest value in Dij . Pick a p oint (x , yj ) in this rectangle. Then i we can write mij xy f (x , yj )xy Mij over all i and j to obtain m n m n f (x , yj )xy i j=1 i=1 j=1 i=1 m n mij xy j=1 i=1 Mij xy.
43 DOUBLE INTEGRALS We call L= j=1 i=1 m n 439 mij xy the lower Riemann sum and m n U= j=1 i=1 Mij xy the upper Riemann sum. If m,n lim L = lim U m,n then we write m n f (x , yj )xy i j=1 i=1 f (x, y)dxdy = lim R m,n and we call R f (x, y)dxdy the double integral of f over the rectangle R. The us e of the word double will be justi ed below. Double Integral as Volume Under a Surfa ce Just as the de nite integral of a positive one variable function can be interpr eted as area, so the double integral of a positive two variable function can be interpreted as a volume. Let f (x, y) > 0 with surface S shown in Figure 43.2(I) . Partition the rectangle R as above. Over each rectangle Dij we will construct a box whose height is given by f (x , yj ) as shown in Figure 43.2 (II). Each of the boxes i has a base area of xy and a height of f (x , yj ) so the volume of each i of these boxes is f (xi , yj )xy. So the volume under the surface S is then app roximately, m n f (x , yj )xy i j=1 i=1 V As the number of subdivisions grows, the tops of the boxes approximate the surfa ce better, and the volume of the boxes gets closer and closer to the volume unde r the graph of the function. Thus, we have the following result, If f (x, y) > 0 then the volume under the graph of f above the region R is f (x, y)dxdy R
440 APPENDIX Figure 43.2 Double Integral as Area If we choose a function f (x, y) = 1 everywh ere in R then our integral becomes: Area of R = R 1dxdy That is, when f (x, y) = 1, the integral gives us the area of the region we are integrating over. Example 43.1 Use the Riemann sum with n = 3, m = 2 and sample point the upper right corner of each subrectangle to estimate the volume under z = xy and above the rectangle 0 x 6, 0 y 4. Solution. The interval on the xaxis i s to be divided into n = 3 subintervals of equal length, so x = 60 = 2. Likewise, the interval on the yaxis is to be divided 3 into m = 2 subintervals, also with w idth y = 2; and the rectangle is divided into six squares with sides 2. Next, the upper right corners are at (2, 2), (4, 2) and (6, 2), for the lower three squar es, and (2, 4), (4, 4) and (6, 4), for the upper three squares. The approximatio n is then [f (2, 2) + f (4, 2) + f (6, 2))+f (2, 4) + f (4, 4) + f (6, 4)] 2 2 = [4 + 8 + 12 + 8 + 16 + 24] 4 = 72 4 = 288
43 DOUBLE INTEGRALS 441 Example 43.2 Values of f (x, y) are given in the table below. Let R be the recta ngle 0 x 1.2, 2 y 2.4. Find Riemann sums which are reasonable over and under es timates for R f (x, y)dxdy with x = 0.1 and y = 0.2 y\ x 2.0 2.2 2.4 1.0 5 4 3 1.1 7 6 5 1.2 10 8 4 Solution. We mark the values of the function on the plane, as shown in Figure 43 .3, so that we can guess respectively at the smallest and largest value the func tion takes on each subrectangle. Figure 43.3 Lower sum = (4 + 6 + 3 + 4)xy = (17)(0.1)(0.2) = 0.34 Upper sum = (7 + 10 + 6 + 8)xy = (31)(0.1)(0.2) = 0.62 Integral Over Bounded Regions That Are Not Rectangles The region of integration R can be of any bounded shape not just rect angles. In our presentation above we chose a rectangular shaped region for conve nience since it makes the summation limits and partitioning of the xyplane into s quares or rectangles simpler. However, this need not be the case. We can instead picture covering an arbitrary shaped region in the xyplane with rectangles so th at either all the rectangles lie just inside the region or the rectangles extend just outside the region (so that the region is contained inside our rectangles) as shown in Figure 43.4. We can then compute either the
442 APPENDIX minimum or maximum value of the function on each rectangle and compute the volum e of the boxes, and sum. Figure 43.4 The Average of f (x, y) As in the case of single variable calculus, the average value of f (x, y) over a region R is de ned by 1 Area of R f (x, y)dxd y. R Iterated Integrals In the previous discussion we used Riemann sums to approximat e a double integral. In this section, we see how to compute double integrals exa ctly using one variable integrals. Going back to our de nition of the integral ove r a region as the limit of a double Riemann sum: m n f (x , yj )xy i j=1 i=1 m n f (x , yj )x y i j=1 m i=1 n f (x, y)dxdy = lim R m,n = lim m,n = lim = lim We now let m,n y i=1 b f (x , yj )x i j=1 m m y j=1 b a f (x, yj )dx F (yj ) = a f (x, yj )dx
43 DOUBLE INTEGRALS and, substituting into the expression above, we obtain m d F (yj )y j=1 d b 443 f (x, y)dxdy = lim R m = c F (y)dy = c a f (x, y)dxdy. Thus, if f is continuous over a rectangle R then the integral of f over R can be expressed as an iterated integral. To evaluate this iterated integral, rst perfo rm the inside integral with respect to x, holding y constant, then integrate the result with respect to y. Example 43.3 16 8 Compute 0 0 12 Solution. We have 16 0 0 8 x 4 y 8 dxdy. 12 x y dxdy = 4 8 = 16 0 16 0 16 0 8 12 x y dx dy 4 8 8 x2 xy 12x 8 8 dy 0 16 = 0 y2 (88 y)dy = 88y 2 = 1280 0
We note, that we can repeat the argument above for establishing the iterated int egral, reversing the order of the summation so that we sum over j rst and i secon d (i.e. integrate over y rst and x second) so the result has the order of integra tion reversed. That is we can show that b d f (x, y)dxdy = R a c f (x, y)dydx. Example 43.4 8 16 Compute 0 0 12 x 4 y 8 dydx.
444 Solution. We have 8 0 0 16 APPENDIX 12 x y dydx = 4 8 = 8 0 8 0 8 0 16 12 x y dy dx 4 8 16 xy y 2 12y 4 16 dx 0 8 0 = 0 (176 4x)dx = 176x 2x2 = 1280 Iterated Integrals Over Non Rectangular Regions So far we looked at double integ rals over rectangular regions. The problem with this is that most of the regions are not rectangular so we need to now look at the following double integral, f (x, y)dxdy R where R is any region. We consider the two types of regions shown in Figure 43.5 . Figure 43.5 In Case 1, the iterated integral of f over R is de ned by b g2 (x) f (x, y)dxdy = R a g1 (x) f (x, y)dydx
43 DOUBLE INTEGRALS 445 This means, that we are integrating using vertical strips from g1 (x) to g2 (x) and moving these strips from x = a to x = b. In case 2, we have d h2 (y) f (x, y)dxdy = R c h1 (y) f (x, y)dxdy so we use horizontal strips from h1 (y) to h2 (y). Note that in both cases, the limits on the outer integral must always be constants. Remark 43.1 Chosing the o rder of integration will depend on the problem and is usually determined by the function being integrated and the shape of the region R. The order of integratio n which results in the simplest evaluation of the integrals is the one that is pre ferred. Example 43.5 Let f (x, y) = xy. Integrate f (x, y) for the triangular re gion bounded by the xaxis, the yaxis, and the line y = 2 2x. Solution. Figure 43.6 shows the region of integration for this example. Figure 43.6 Graphically integrating over y rst is equivalent to moving along the x axis from 0 to 1 and integrating from y = 0 to y = 2 2x. That is, summing up
446 the vertical strips as shown in Figure 43.7(I). 1 22x APPENDIX xydxdy = R 0 1 0 xydydx xy 2 2 1 22x 0 2 = 0 1 dx = 2 3 1 x(2 2x)2 dx 0 1 =2 0 x2 2 3 x4 (x 2x + x )dx = 2 x + 2 3 4 = 0 1 6 If we choose to do the integral in the opposite order, then we need the y = 2 2x i.e. express x as function of y. In this case we get egrating in this order corresponds to integrating from y = 0 to y = zontal strips ranging from x = 0 to x = 1 1 y, as shown in Figure 2 1 1 y 2 xydxdy = R 0 2 0 xydxdy x2 y 2 2 0 1 1 2 y = 0 0 2 1 dy = 2 3 2 0 1 y(1 y)2 dy 2 2 1 = 2 to invert 1 x = 1 2 y. Int 2 along hori 2 43.7(II)
y y2 y3 y4 (y y + )dy = + 4 4 6 32 = 0 1 6 Figure 43.7
43 DOUBLE INTEGRALS 447 Example 43.6 Find R (4xy y 3 )dxdy where R is the region bounded by the curves y = x and y = x3 . Solution. A sketch of R is given in Figure 43.8. Using horizon tal strips we can write 1 3 y2 1 y (4xy y )dxdy = R 0 3 (4xy y 3 )dxdy 2x2 y xy 3 0 3 y2 y 1 = dy = 0 1 2y 3 y 3 y 5 dy 55 156 5 10 3 8 3 13 1 = y 3 y 3 y6 4 13 6 = 0 Figure 43.8 Example 43.7 Sketch the region of integration of 2 0 4x2 4x2 xydydx Solution. A sketch of the region is given in Figure 43.9.
448 APPENDIX Figure 43.9
43 DOUBLE INTEGRALS Practice Problems 449 Problem 43.1 Set up a double integral of f (x, y) over the region given by 0 < x < 1; x < y < x + 1. Problem 43.2 Set up a double integral of f (x, y) over the part of the unit square 0 x 1; 0 y 1, on which y x . 2 Problem 43.3 Set up a dou ble integral of f (x, y) over the part of the unit square on which both x and y are greater than 0.5. Problem 43.4 Set up a double integral of f (x, y) over the part of the unit square on which at least one of x and y is greater than 0.5. P roblem 43.5 Set up a double integral of f (x, y) over the part of the region giv en by 0 < x < 50 y < 50 on which both x and y are greater than 20. Problem 43.6 Set up a double integral of f (x, y) over the set of all points (x, y) in the rst quadrant with x y 1. Problem 43.7 Evaluate exy dxdy, where R is the region in t he rst quadrant in which R x + y 1. Problem 43.8 Evaluate ex2y dxdy, where R is the region in the rst quadrant in R which x y. Problem 43.9 Evaluate (x2 + y 2 )dxdy , where R is the region 0 x y L. R Problem 43.10 Evaluate f (x, y)dxdy, where R is the region inside the unit square in R which both coordinates x and y are gre ater than 0.5.
450 APPENDIX Problem 43.11 Evaluate (x y + 1)dxdy, where R is the region inside the unit squa re R in which x + y 0.5. Problem 43.12 1 1 Evaluate 0 0 xmax(x, y) y x.

44 DOUBLE INTEGRALS IN POLAR COORDINATES 451 44 Double Integrals in Polar Coor inates There are regions in the plane that are not easily use as omains of iterate i ntegrals in rectangular coor inates. For instance, regions such as a isk, ring, or a portion of a isk or ring. We start by establishing the relationship betwe en Cartesian an polar coor inates. The Cartesian system consists of two rectang ular axes. A point P in this system is uniquely etermine by two points x an y as shown in Figure 44.1(a). The polar coor inate system consists of a point O, calle the pole, an a half axis starting at O an pointing to the right, known as the polar axis. A point P in this system is etermine by two numbers: the i stance r between P an O an an angle between the ray OP and the polar axis as s hown in igure 44.1(b). igure 44.1
The Cartesian and polar coordinates can be combined into one gure as shown in ig ure 44.2. igure 44.2 reveals the relationship between the Cartesian and polar c oordinates: r= x2 + y 2 x = r cos y = r sin y tan = x .

452 APPENDIX
Figure 44.3 Partition the interval into m equal su itervals, and the interval a r into n equal su intervals, thus o taining mn su rectangles as shown in Figur e 44.3( ). Choose a sample point (rij , ij ) in the subrectangle Rij de ned by ri1 r ri and j1 j . Then m n (x, y)dxdy R j=1 i=1
(rij , ij )Rij
igure 44.2 A double integral in polar coordinates can be de ned as ollows. Suppo se we have a region R = {(r, ) : a r b, } as shown in Figure 44.3(a).
44 DOUBLE INTEGRALS IN POLAR COORDINATES 453
where Rij is the area of the subrectangle Rij . To calculate the area of Rij , lo ok at Figure 44.4. If r and are small then Rij is approximately a rectangle with a rea rij r so Rij rij r. Thus, the double integral can be approximated by a Riemann m n (x, y)dxdy R j=1 i=1
Taking the limit as m, n we obtain f (x, y)dxdy = R f (r, )rdrd.
igure 44.4
Example 44.1 2 2 Evaluate R ex +y dxdy where R : x2 + y 2 1.

(rij , ij )rij r
454 Solution. We have e R x2 +y 2 1 1x2 APPENDIX dxdy = = = 0 1 1x2 2 1 r2 0 2 0 ex 2 +y 2 dydx 2 0 e rdrd = 1 r2 d e 2 0 1 1 (e 1)d = (e 1) 2 Exam le 44.2 Com ute the area of a circle of radius 1. Solution. The area is giv en by 1 1 1x2 2 1 1x2 dydx = 0 0 2 rdrd d = 0 1 = 2 Exam le 44.3 Evaluate f (x, y) = 1 x2 +y 2 over the region D shown below. Solution. We have 0 4 2 1 1 rdrd = r2 4 ln 2d = 0 ln 2 4
44 DOUBLE INTEGRALS IN POLAR COORDINATES 455 Exam le 44.4 or each of the regions shown below, decide whether to integrate us ing rectangular or olar coordinates. In each case write an iterated integral of an arbitrary function f (x, y) over the region.
Solution. 2 3 (a) 0 0 f (r, )rdrd 3 2 (b) 1 1 x 2
(x, y)dydx 2 3 (c) 0 1 x1
(x, y)dyd
456 APPENDIX
Answer Keys Section 1 1.1 A = {2, 3, 5} 1.2 (a)S = {T T T, T T H, T HT, T HH, HT T, HT H, HH T, HHH} (b)E = {T T T, T T H, HT T, T HT } (c) = {x : x is an element o S with more than one head} 1.3 E 1.4 E = 1.5 (a) Since every element o A is in A, A A. (b) Since every element in A is in B and every element in B is in A, A = B. (c) If x is in A then x is in B since A B. But B C and this implies that x is in C. Hence, every element of A is also in C. This shows that A C 1.6 The result i s true for n = 1 since 1 = 1(1+1) . Assume that the equality is 2 true for 1, 2, , n. Then
1 + 2 + + n + 1 =(1 + 2 + + n) + n + 1 n(n + 1) n = + n + 1 = (n + 1)[ + 1] 2 2 (n + 1)(n + 2) = 2 1.7 Let Sn = 12 + 22 + 32 + + n2 . For n = 1, we have S1 = 1 = 1(1+1)(2+1) . Sup pose that Sn = n(n+1)(2n+1) . We next want to show that 6 6 Sn+1 = (n+1)(n+2)(2n +3) . Indeed, Sn+1 = 12 + 22 + 32 + + n2 + (n + 1)2 = 6 n(n+1)(2n+1) 6 + (n + 1)2 = (n + 1) n(2n+1) 6 +n+1 = (n+1)(n+2)(2n+3) 6 457
458 ANSWER KEYS 1.8 The result is true for n = 1. Suppose true up to n. Then (1 + x)n+1 =(1 + x) (1 + x)n (1 + x)(1 + nx), since 1 + x > 0 =1 + nx + x + nx2 =1 + nx2 + (n + 1)x 1 + (n + 1)x 1.9 (a) 55 tacos with tomatoes or onions. (b) There are 40 tacos wit h onions. (c) There are 10 tacos with onions but not tomatoes 1.10 (a) Stu ents taking English an history, but not math, are in the brown region; there are 20 of these. (b) Stu ents taking none of the three courses are outsi e the three ci rcles; there are 5 of these. (c) Stu ents taking math, but neither English nor h istory, are in the blue region; there are 11 of these. ( ) Stu ents taking Engli sh, but not history, are in the yellow an gray regions; there are 25 + 17 = 42 of these. (e) Stu ents taking only one of the three subjects are in the yellow, green, an blue regions; there are 25 + 10 + 11 = 46 of these. (f) Stu ents taki ng two of the three subjects are in the pink, gray, an brown regions; there are 9 + 17 + 20 = 46 of these. 1.11 Since S = {(1, 1), (1, 2), (1, 3), (1, 4), (1, 5), (1, 6), (2, H), (2, T ), (3, 1), (3, 2), (3, 3), (3, 4), (3, 5), (3, 6), (4, H), (4, T ), (5, 1), (5, 2) , (5, 3), (5, 4), (5, 5), (5, 6), (6, H), (6, T )} n(S) = 24

459 Section 2 2.1
2.2 Since A B, we have A B = B. Now the result follows from the previous problem . 2.3 Let G = event that a viewer watched gymnastics B = event that a viewer wat ched baseball S = event that a viewer watched soccer Then the event the group tha t watched none of the three sports during the last year is the set (G B S)c 2.4 T he events R1 R2 and B1 B2 represent the events that both ball are the same color and therefore as sets they are disjoint 2.5 880 2.6 50% 2.7 5% 2.8 60 2.9 53% 2 .10 Using Theorem 2.3, we nd n(A B C) =n(A (B C)) =n(A) + n(B C) n(A (B C) + (n(B) + n(C) n(B C)) n((A B) (A C)) =n(A) + (n(B) + n(C) n(B C)) (n(A C) n(A B C)) =n(A) + n(B) + n(C) n(A B) n(A C) n(B C) + n(A B C)
460 ANSWER KEYS
2.11 50 2.12 10 2.13 (a) 3 (b) 6 2.14 32 2.15 20% 2.16 (a) Let x A (B C). Then x A and x B C. Thus, x A and (x B or x C). This implies that (x A and x B) o A and x C). Hence, x A B or x A C, i.e. x (A B) (A C). The converse is si (b) Let x A (B C). Then x A or x B C. Thus, x A or (x B and x C). This im hat (x A or x B) and (x A or x C). Hence, x A B and x A C, i.e. x (A B) he converse is similar. 2.17 (a) B A (b) A B = or A B c . (c) A B A B (d) (A Section 3 3.1 (a) 100 (b) 900 (c) 5040 (d) 90,000 3.2 (a) 336 (b) 6 3.3 6 3.4 9 0 3.5 3.6 36 3.7 380 3.8 6,497,400 3.9 5040 3.10 3840
461 Section 4 4.1 m = 9 and n = 3 4.2 (a) 456976 (b) 358800 4.3 (a) 15600000 (b) 112 32000 4.4 (a) 64000 (b) 59280 4.5 (a) 479001600 (b) 604800 4.6 (a) 5 (b) 20 (c) 60 (d) 120 4.7 20 4.8 (a) 362880 (b) 15600 4.9 m = 13 and n = 1 or n = 12 4.10 1 1480 4.11 300 4.12 10 4.13 28 4.14 4060 m! 4.15 Recall that P (m, n) = (mn)! = n! C(m, n). Since n! 1, we can multiply both si es by C(m, n) to obtain P (m, n) = n!C(m, n) C(m, n) 4.16 125970 4.17 43,589,145,600 4.18 144 Section 5 5.1 19,958, 400 5.2 9,777,287,520 5.3 5.3645 1028 5.4 4,504,501 5.5 12,600 5.6 1260 5.7 1771 , 10,626 5.8 21 5.9 (a) 28 (b) 60 (c) 729 5.10 211876 5.11 36,052,387,482,172,42 5 5.12 (a) 136 (b) C(15, 2)C(8, 2) Section 6
462 6.1 (a) S = {1, 2, 3, 4, 5, 6} (b) {2, 4, 6} 6.2 ANSWER KEYS {(1, 1), (1, 2), (1, 3), (1, 4), (1, 5), (1, 6), (2, H), (2, T ), (3, 1), (3, 2) , (3, 3), (3, 4), (3, 5), (3, 6), (4, H), (4, T ), (5, 1), (5, 2), (5, 3), (5, 4 ), (5, 5), (5, 6), (6, H), (6, T )} 6.3 50% 6.4 (a) S = {(1, 1), (1, 2), (1, 3), (1, 4), (2, 1), (2, 2), (2, 3), (2, 4), (3, 1), (3, 2), (3, 3), (3, 4), (4, 1), (4, 2), (4, 3), (4, 4)} (b) E c = {(1, 1), (1, 2), (2, 1)} (c) 0.625 ( ) 0.75 ( e) 0.625 6.5 (a) 0.48 (b) 0 (c) 1 ( ) 0.36 (e) 0.04 6.6 (a) 0.375 (b) 0.25 (c) 0 .5 ( ) 0 (e) 0.375 (f) 0.125 6.7 (a) 6.3 1012 (b) 0.041 6.8 25% 6.9 12.5% 5 4 6.1 0 (a) n2 (b) n2 6.11 1 6.6 1014 6.12 (a) 10 (b) 40% 6.13 (a) S = {D1 D2 , D1 N1 , D1 N2 , D1 N3 , D2 N1 , D2 N2 , D2 N3 , N1 N2 , N1 N3 , N2 N3 } (b) 0.1 Section 7 7.1 (a) 0.78 (b) 0.57 (c) 0 7.2 0.32 7.3 0.308 7.4 0.555 7.5 Since P (A B) 1, we have P (A B) 1. A P (A) + P (B) to both si es to obtain P (A) + P (B) P (A ) P (A) + P (B) 1. But the left han si e is just P (A B). 7.6 (a) 0.181 (b) 0.8 18 (c) 0.545 7.7 0.889 7.8 No 7.9 0.52 7.10 0.05 7.11 0.6 7.12 0.48 7.13 0.04

463 7.14 0.09 7.15 0.5 Section 8 8.1 8.2 8.3 8.4 8.5 8.6 8.7 P (A) = 0.6, P (B) = 0.3, P (C) = 0.1 0.1875 0.444 0.167 3 2 The probability is 5 4 + 2 3 = 3 = 0.6 5 4 5
464 ANSWER KEYS 8.8 The probability is 3 5 2+2 5 5 3 5 = 12 25 = 0.48 8.9 0.14 36 8.10 65 Section 9 9.1 0.173 9.2 0.205 9.3 0.467 9.4 0.5 9.5 (a) 0.19 (b) 0.60 (c) 0.31 (d) 0.317 (e) 0.613 9.6 0.151 9.7 0.133 9.8 0.978 9.9 0.986 7 9.10 1912 1 1 9.11 (a) 221 (b) 169 1 9.12 114 9.13 80.2% 9.14 (a) 0.021 (b) 0.2 381, 0.2857, 0.476 9.15 (a) 0.48 (b) 0.625 (c) 0.1923
465 Section 10 6 10.1 (a) 0.26 (b) 13 10.2 0.1584 10.3 0.0141 10.4 0.29 10.5 0.4 2 10.6 0.22 10.7 0.657 10.8 0.4 10.9 0.45 10.10 0.66 15 10.11 (a) 0.22 (b) 22 4 10.12 (a) 0.56 (b) 7 10.13 0.36 1 10.14 3 10.15 0.4839 7 17 10.16 (a) 140 (b) 17 Section 11 11.1 (a) Dependent (b) Independent 11.2 0.02 11.3 (a) 21.3% (b) 21.7 % 11.4 0.72 11.5 4 11.6 0.328 11.7 0.4 11.8 We have 1 1 1 = = P (A)P (B) 4 2 2 1 1 1 P (A C) =P ({1}) = = = P (A)P (C) 4 2 2 1 1 1 P (B C) =P ({1}) = = = P (B)P (C) 4 2 2 P (A B) =P ({1}) = It follows that the events A, B, and C are pairwis e independent. However, P (A B C) = P ({1}) = 1 1 = = P (A)P (B)P (C). 4 8
466 ANSWER KEYS
Thus, the events A, B, and C are not indepedent 11.9 P (C) = 0.56, P (D) = 0.38 11.10 0.93 11.11 0.43 (ABC) 11.12 (a) We have P (A B C) = P P (BC) = P (A)P (B)P (C ) = P (A). Thus, P (B)P (C) A and B C are independent. (b) We have P (A BC) = P ( A(BC)) = P ((AB)(AC)) = P (AB)+P (AC)P (ABC) = P (BC) P (BC) P (B)+P (C)P (BC (B)P (C c )+P P (A)P (B)+P (A)P (C)P (A)P (B)P (C) = P (A)P (B)[1P (C)]+P (A)P (C) = P (A)P (B)P (C c )+P (A)P (C) P (A)+P (B)P (A)P (B) P (B)+P (C)P (B)P (C) P (C) P (A)[P (B)P (C c )+P (C)] = P (A). Hence, A and B C are independent P (B)P (C c )+P (C) = 11.13 (a) We have S = {HHH, HHT, HT H, HT T, T HH, T HT, T T H, T T T } A = {HHH , HHT, HT H, T HH, T T H} B = {HHH, HHT, T HT, T T T } (b) P (A) = 0.5(c) 4 (d) We have B C = {HHH, HT H}, so P (B C) = 1 . 5 4 That is equal to P (B)P (C), so B and C are independent 11.14 0.65 11.15 (a) 0.70 (b) 0.06 (c) 0.24 (d) 0.72 (e) 0.4615 Section 12 12.1 15:1 12.2 62.5% 12.3 1:1 12.4 4:6 12.5 4% 12.6 (a) 1:5 ( b) 1:1 (c) 1:0 (d) 0:1 12.7 1:3 12.8 (a) 43% (b) 0.3 2 12.9 3 Section 13 13.1 (a ) Continuous (b) Discrete (c) Discrete (d) Continuous 13.2 If B and G stand for brown and green, the probabilities for BB, BG, GB, 5 5 15 3 and GG are, respecti velym 8 4 = 14 , 5 3 = 15 , 3 5 = 56 , and 3 2 = 28 . 7 8 7 56 8 7 8 7 The resul ts are shown in the following table.
467 Element of sample space Probability x 5 2 BB 14 15 1 BG 56 15 GB 1 56 3 GG 0 28 13.3 13.4 13.5 13.6 13.7 13.8 13.9 0.139 0.85 1 n 2 0.5 0.4 0.9722 (a) s {(N S, N S, N S)} 0 1 s {(S, N S, N S), (N S, S, N S), (N , N S, S)} X(s) = s {(S, S, N S), (S, N S, S), (N S, S, S)} 2 3 s {(S, S, S)} (b) 0.09 (c) 0.36 (d) 0.41 (e) 0.14 1 1 13.10 P (X = 0) = 2 , P (X = 1) = 6 , P (X = 2) = 1 13.11 1+e 13.12 0.267 Section 14 14.1 (a) x 0 p(x) 1 8 (b) 1 3 8 1 , 12 P (X = 3) = 1 4 2 3 8 3 1 8
468 14.2
x<0
0,
, 0x<1
8 1 , 1x<2 F (x) = 2
, 2x<3
1, 3x
ANSWER KEYS 14.3 x 2 1 p(x) 36 3 2 36 4 3 36 5 4 36 6 5 36 7 6 36 8 5 36 9 4 36 10 11 12 3 36 2 36 1 36 14.4 X 1 2 3 4 5 6 Event F (P,f) (P,P,F) (P,P,P,F) (P,P,P,P,F) (P,P,P,P,P,F) p(x) 1 6 5 6 5 6 5 6 5 6 5 6 1 5 4 5 4 5 4 5 4 5 =1 6 1 1=6 4 1 33=1 4 6 2 331= 4 2 2 331= 4 2 1 6 1 6
469 14.5 x 0 1 2 p(x) 1/4 1/2 1/4 14.6 n F (n) =P (X n) = k=0 n P (X = k) = k=0 1 3 2 3 k 11 2 3 = 3 1 =1 2 3 n+1 2 3 n+1 14.7 (a) For n = 2, 3, , 96 we have P (X = n) = and P (X = 1) = (b) 5 100 95 94 95 n + 2 5 100 99 100 n + 2 100 n + 1 P (Y = n) = 5 n 95 10 n 100 10 , n = 0, 1, 2, 3, 4, 5
470 14.8 p(x) = 14.9 P (X = 2) =P (RR) + P (BB) = = and P (X = 1) = 1 P (X = 2) = 1 14.10 (a) p(0) = p(1) =3 p(2) =3 p(3) = (b) 2 3 1 3 1 3 1 3 3 2 3 ANSWER KEYS
3 10 4 10 3 10 x = 4 x=1 x=4 C(3, 2) C(4, 2) + C(7, 2) C(7, 2) 3 6 9 3 + = = 21 21 21 7 3 4 = 7 7 2 3 2 3 2 1 2 3 4 5 6 14.11 p(2) = 36 , p(3) = 36 , p(4) = 36 , p(5) = 36 , p(6) = 36 , p( 7) = 36 , p(8) = 5 4 3 2 1 , p(9) = 36 , p(10) = 36 , p(11) = 36 , and p(12) = 3 6 and 0 otherwise 36
471 14.12 C(3, 0)C(12, 3) C(15, 3) C(3, 1)C(12, 1) p(1) = C(15, 3) C(3, 2)C(12, 1) p(2) = C(15, 3) C(3, 3)C(12, 0) p(3) = C(15, 3) p(0) = 220 455 198 = 455 36 = 455 1 = 455 = Section 15 15.1 7 15.2 $ 16.67 15.3 E(X) = 10 1 2 5 = 0 Therefore, you should co me out about even 6 6 if you play for a long time 15.4 1 15.5 E(X) = $0.125 So the owner will make on average 12.5 cents per spin 15.6 $26 15.7 $110 15.8 0.54 15.9 897 15.10 (a) 0.267 (b) 0.449 (c) 1.067 15.11 (a) 0.3 (b) 0.7 (c) p(0) = 0.3, p (1) = 0, p(P (X 0)) = 0 (d) 0.2 x = 2 0.3 x=0 0.1 x = 2.2 p(x) = x=3 0 otherwise (e) 1.12 15.12 (a) 2400 (b) Since E(V ) < 2500 the answer is no
472 15.13 (a) C(3, 3)C(7, 1) 210 C(3, 2)C(7, 2) p(2) =P (X = 2) = 210 C(3, 1)C(7 , 3) p(3) =P (X = 3) = 210 C(3, 0)C(7, 4) p(4) =P (X = 4) = 210 p(1) =P (X = 1) = (b) x<1 0 7 210 1 x < 2 70 2x<3 F (x) = 210 175 3 x < 4 210 16 1 16.1 (a) c = 30 (b) 3.333 (c) 8.467 1 3 16.2 (a) c = 9 (b) p(1) = 2 , p(1) = 9 , p(2) = 9 E(X 2 ) = 7 3 16.3 (a) x x = 1, 2, 3, 4, 5, 6 21 p(x) = 0 otherwi se 4 (b) 7 (c) E(X) = 4.333 16.4 Let D enote the range of X. Then ANSWER KEYS 7 210 63 = 210 105 = 210 35 = 210 = 4 9 (c) E(X) = 1 an
E(aX 2 + bX + c) = xD (ax2 + bx + c)p(x) ax2 p(x) + xD xD = =a bxp(x) + xD cp(x) p(x) xD x p(x) + b xD 2 xD 2 xp(x) + c =aE(X ) + bE(X) + c
473 16.5 16.6 16.7 16.8 16.9 0.62 $220 0.0314 0.24 (a) P (X = 2) =P (RR) + P (BB ) = = an P (X = 1) = 1 P (X = 2) = 1 (b) E(2X ) = 2 16.10 (a) P = 3C + 8A + 5S 3 00 (b) $1,101 16.11 We have 1 1 E(2 ) = (2 ) 1 + (22 ) 2 + = 2 2 X 1 C(3, 2) C(4, 2) + C(7, 2) C(7, 2) 3 6 9 3 + = = 21 21 21 7 4 3 = 7 7 1 n=1 The series on the right is divergent so that E(2X ) does not exist Section 17 17 .1 0.45 17.2 374 17.3 (a) c = 17.4 (a) 1 55 (b) E(X) = 1.09 Var(X) = 1.064 C(3, 3)C(7, 1) C(10, 4) C(3, 2)C(7, 2) p(2) = C(10 , 4) C(3, 1)C(7, 3) p(3) = C(10, 4) C(3, 0)C(7, 4) p(4) = C(10, 4) p(1) = 1 30 3 = 10 1 = 2 1 = 6 = (b) E(X) = 2.8 Var(X) = 0.56 1 17.5 (a) c = 30 (b) E(X) = 3.33 and E(X(X 1)) = 8 .467 (c) E(X 2 ) =
474 11.797 and Var(X) = 0.7081 17.6 E(Y ) = 21 and Var(Y ) = 144 17.7 (a) STA200 STA291 (b) (c) 0.24 0.42 ANSWER KEYS ADD 0.42 0.16 0.58 SWITCH 0.18 0.24 0.42 0.60 0.40 1.00 = 0.5714 x 15 12 8 7 p(x) 0.42 0.18 0.16 0.24 (d) E(X) = 11.42 and Var(X) = 12.0036 17.8 (a) P (X = 2) =P (RR) + P (BB) = = an d P (X = 1) = 1 P (X = 2) = 1 2 (b) E(X) = 7 and E(X 2 ) = 16 (c) Var(X) = 108 7 49 4 17.9 E(X) = 10 ; Var(X) = 9.86; SD(X) = 3.137 17.10 E(X) = p Var(X) == p(1 p) C(3, 2) C(4, 2) + C(7, 2) C(7, 2) 6 9 3 3 + = = 21 21 21 7 3 4 = 7 7 Section 18 18.1 0.3826 18.2 0.211 18.3 1 8 , if x = 0, 3 3 , if x = 1, 2 p(x) = 8 0, otherwise
18.4 0.096 18.5 $60 18.6 0.0488
475 18.7 0.925 18.8 0.144 18.9 (a) 0.1875 (b) 0.5 18.10 0.1198 18.11 0.6242 18.1 2 1.82284263 105 18.13 0.63 18.14 154 18.15 0.135 18.16 0.784 18.17 (a) 0.2321 (b ) 0.2649 (c) 0.1238 18.18 0.2639 18.19 $ 985 18.20 Note that E(S) = 100 + 50E(Y ) 10E(Y 2 ). n 1 2 3 E(Y ) 0.20 0.40 0.60 E(Y 2 ) 0.20 0.48 0.84 E(S) 100+10 2 = 108 100+20 4.8 = 115.2 100+30 8.4 = 121.6 where E(Y 2 ) = n(n 1)p2 + np. Section 19 19.1 0.0027 19.2 0.1251 19.3 3.06 107 1 9.4 (a) 0.577 (b) 0.05 19.5 (a) 0.947 (b) 0.762 (c) 0.161 19.6 0.761897 19.7 (a) 0.5654 (b) 0.4963 19.8 0.4232 19.9 2 19.10 $7231.30 19.11 699 19.12 0.1550 19.1 3 0.7586 19.14 4 19.15 (a) 0.2873 (b) mean = 20 and variance = 4.47 19.16 0.0838
476 19.17 (a) 0.0821 (b) 0.1268 ANSWER KEYS Section 20 9 k1 1 20.1 (a) 0.1 (b) 0.09 (c) 10 10 20.2 0.387 20.3 0.916 20.4 (a) 0.001999 (b) 1000 20.5 (a) 0.046875 (b) 0.0039 20.6 0.359546 49 1 2 2 (b) 3 20.7 (a) 3 3 and 1 2 3 3 x 20.8 (a) pX (x) = (0.85)(0.15) , x = 0, 1, 2, and 0 other wise (b) pY (y) = (0.85)(0.15)y1 , y = 1, 2, 3, and 0 otherwise. Thus, Y is a geo metric random variable with parameter 0.85 20.9 (a) 0.1198 (b) 0.3999 20.10 We h ave P (X > i + j X > i) = P (X > i + j, X > i) P (X > i) P (X > i + j) (1 p)i+j = = P (X > i) (1 p)i =(1 p)j = P (X > j) 20.11 0.053 20.12 (a) 10 (b) 0.81 20.13 (a) X is a geometric distribution with p mf p(x) = 0.4(0.6)x1 , x = 1, 2, (b) X is a binomial random variable with pmf p(x ) = C(20, x)(0.60)x (0.40)20x where x = 0, 1, , 20 20.14 0.1728 20.15 0.0307 4 1 20.16 (a) X is negative binomial distribution with r = 3 and p = 52 = 13 . 1 3 1 2 k3 So p(k) = C(k 1, 2) 13 (b) 0.018 13 20.17 E(X) = 24 and Var(X) = 120 20.18 0 .109375 20.19 0.1875 20.20 0.2898 20.21 0.022 20.22 (a) 0.1198 (b) 0.0254 20.23 20
477 20.24 20.25 20.26 20.27 20.28 0.0645 C(t 1, 2)(1/6)3 (5/6)t3 0.32513 0.1988 k 0 1 2 3 4 P(X=k) 0.468 0.401 0.117 0.014 7.06 104 5 1.22 105 6 4.36 10 8 20.29 20.30 20.31 20.32 20.33 20.34 20.35 20.36 20.37 20.38 0.247678 0.073 (a) 0.214 (b) E(X) = 3 and Var(X) = 0.429 0.793 C(2477,3)C(121373,97) C(123850,100) 0.033 0.2880 4 0.375 0.956 Section 21 21.1 (a) C(2, 2) = 0.1 C(5, 2) C(2, 1)C(2, 1) = 0.4 p(6) = C(5, 2) C( 2, 2) p(10) = = 0.1 C(5, 2) C(1, 1)C(2, 1) p(11) = = 0.2 C(5, 2) C(1, 1)C(2, 1) p(15) = = 0.2 C(5, 2) p(2) =
478 (b) x<2 0 5 (c) 8.8 21.2 ANSWER KEYS
0.1 2 x < 6
0.5 6 x < 10 F (x) =
0.6 10 x < 11
x (, 0) [0, 1) [1, 2) [2, 3) [3, ) P (X x) 0 0.495 0.909 0.996 1 21.3 (a) P (X = 1) =P (X 1) P (X < 1) = F (1) lim F n 1 1 n 1 1 1 = = 2 4 4 P (X = 2) =P (X 2) P (X < 2) = F (2) lim F n 2 1 n = 11 12 1 21 2 4 = 1 6 n P (X = 3) =P (X 3) P (X < 3) = F (3) lim F =1 (b) 0.5 21.4 P (X = 0) =F (1) = 1 2 1 11 = 12 12 3 1 n 3 1 1 = 5 2 10 4 3 1 P (X = 2) =F (2) F (2 ) = = 5 5 5 9 4 1 P (X = 3) =F (3) F ( 3 ) = = 10 5 10 9 1 P (X = 3.5) =F (3.5) F (3.5 ) = 1 = 10 10 P (X = 1) =F (1) F 1 ) =
479 an 0 otherwise. 21.5 (a) 0.1 x = 2 0.2 x = 1.1 0.3 x=2 p(x) = 0.4 x=3 rwise (b) 0 (c) 0.4 ( ) 0.444 21.6 (a) 0.1 (b) x < 1.9 0 0.1 1.9 x < 0.1 2 F (x) = 0.5 2x<3 0.6 3x<4 1 x4 The graph of F (x) is shown below. (c) F (0) = 0.2; F (2) = 0.5; F (F (3.1)) = 0.2. ( ) 0.5 (e) 0.64 21.7 (a) = 4 0.4 x=1 p(x) = x=4 0.3 0 otherwise
0.3 x
480 (b)E(X) = 0.4, Var(X) = 9.84, an SD(X) = 3.137 21.8 (a) 1 12 2 12 ANSWER KEYS p(x) = 0 x = 1, 3, 5, 8, 10, 12 x = 2, 4, 6 otherwise (b) F (x) = 12 7 12 9 12 10 12 11 12 1 (c) P (X < 4)0.333. This is not the same as F (4) which is the probability that X 4. The i erence between them is the probability that X is EQUAL to 4. 21.9 (a) We have P (X = 0) =(jump in F (x) at x = 0) = 0 1 1 1 P (X = 1) =(jump in F (x) at x = 1 ) = = 2 4 4 3 1 P (X = 2) =(jump in F (x) at x = 2) = 1 = 4 4 7 3 (b) 16 (c) 16 ( ) 3 8 21.10 (a) 0.125 (b) 0.584 (c) 0.5 ( ) 0.25 (e) 0 1 12 3 12 4 12 6
x<1 1
481 Section 22 22.1 2 22.2 (a) 0.135 (b) 0.233 (c) F (x) = 22.3 k = 0.003, 0.027 22. 4 0.938 22.5 (a) Yes (b) x<0 0 1/2 0 x < 1 f (x) = F (x) = 1/6 1 x < 4 (a) 1 (b) 0.736 22.7(a) f (x) = F (x) = 22.8 0.369 1 22.9 9 22.10 0.469 22.11 0. 132 22.12 0.578 22.13 0.3 511 22.14 512 22.15 2 ex (1+ex )2 1 e 5 x 0 0 x<0 x (b) 0.231 0
482 22.16 The c f of claims pai is H(x) =P (X x X > a) P (X x, X > a) = P (X > a) P (a X x) = P (X > a) = 0 F (x)F (a) 1F (a) ANSWER KEYS xa x>a The cumulative distribution function of claims not paid is H(x) =P (X x X a) P ( X x, X a) = P (X a) = = 22.17 (a) e0.5 (b) e0.4 e0.8 (c) P (Xx) P (Xa) 1 F (x) F (a) xa x>a 1 xa x>a e0.5 e0.8 e0.4 e0.8 Section 23 23.1 (a) 1.2 (b) The cdf is given by 0 x 1 0.2 + 0.2x 1 < x 0 F (x) = 2 0<x1 0.2 + 0.2x + 0.6x 1 x>1 (c) 0.25 (d) 0.4 3 23.2 (a) a = 5 and b = 6 . 5 (b) x
F (x) = f (u)du =
x 1 (3 + 5u2 )du = 3 x 5 5 1 1 (3 + 6u2 )du = 0 5 x 0du = 0 x<0 + 2 x3 0 x 1 5 1 x>1
483 23.3 (a) 4 (b) 0 (c) 23.4 2 9 1 23.5 (a) E(X) = 3 and Var(X) = 2 . SD = 0.471 (b) 0.938 9 23.6 0.5 23.7 .06 7 23.8 E(Y ) = 3 and Var(Y ) = 0.756 23.9 1.867 2 3.10 $328 23.11 0.5 23.12 123.8 23.13 0.139 23.14 2.227 23.15 2694 23.16 a + 2 l n 2 23.17 3 ln 2 23.18 2 23.19 1.05 23.20 6 23.21 0.8409 23.22 (a) fX (x) = 4 x 100 100 3 = 4 3 x 106 for 0 x 100 and 0 elsewhere. (b) 80 Section 24 24.1 (a) The pdf is given by 3x7 0 otherwise 1 4 f (x) = (b) 0 (c) 0.5 24.2 (a) The pdf is given by 1 10 f (x) = 0 5 x 15 otherwise
484 (b) 0.3 (c) E(X) = 10 and Var(X) = 8.33 24.3 (a) x<0 0 x 0x1 F (x) = 1 x>1 (b) P (a X a + b) = F (a + b) F (a) = a + b a = b 1 24.4 n+1 24.5 0.693 24.6 0.667 24.7 500 24.8 1.707 24.9 403.44 2 1 1 24.10 (a) e 2e (b) 1 2e2 2 24.11 3 7 24.12 10 1 24.13 min{1, a } Section 25 25.1 (a) 0.2389 (b) 0.1423 (c) 0.6188 (d) 88 25 .2 (a) 0.7517 (b) 0.8926 (c) 0.0238 25.3 (a) 0.5 (b) 0.9876 25.4 (a) 0.4772 (b) 0.004 25.5 2% 25.6 (a) 0.9452 (b) 0.8186 25.7 (a) 0.2578 (b) 0.832 25.8 0.224 25 .9 75 1 25.10 (a) 302 e2 (b) 1099 25.11 0.4721 (b) 0.1389 (c) 0.6664 (d) 0.58 25.12 0.01287 25.13 0.86 25.14 (a) 0.1736 (b) 1071.20 25.15 0.1151 25.16 (a) 0.1056 ( b) 362.84 25.17 1:18 m 25.18 0.1788 25.19 (a) 0.2514 (b) 0.7496 (c) 16.92 millia m eres ANSWER KEYS
485 Section 26 26.1 0.593 26.2 26.3 0.393 26.4 0.1175 26.5 0.549 26.6 (a) 0.189 (b) 0.250 26.7 (a) 0.1353 (b) 0 .167 (c) 0.05 26.8 0.134 26.9 (a) 0.186 (b) 0.015 26.10 0.435 26.11 10256 26.12 3.540 26.13 5644.23 26.14 173.29 26.15 0.420 26.16 (ln42)2 26.17 e5 510 10! Section 27 27.1 0.124
486 27.2 27.3 27.4 27.5 27.6 27.7
ANSWER KEYS E(X) = 1.5 and Var(X) = 0.75 0.0948 0.419 0.014 480 or t 0 we have X 2 (t) = P (X 2 t) = P ( t < X < t) = ( t) ( t) 1 1 fX 2 (t) = ( t) + 1 1 1 1 = ( t) = t 2 e 2 t 2 Now, taking the derivative (and using the chain rule) we nd
which is the density function of gamma distribution with = = 27.8 E(etX ) = t , t < 27.9 We have 2 ex (x)2 f (x) = (x + 1). () 1 2 1 Thus, the only critic l point of f (x) is x = ( 1). One c n e sily show 1 th t f ( 1) < 0 27.10 100 27.11 ( ) The density function is f (x) = 432 x e 6 x2 0 0x1 elsewhere E(X) = 18 nd = 10.39 (b) 1313 27.12 (a) 60 (b) E(X) = 0.571 and Var(X) = 0.031 27.13 (a) We have x F (x) = 0 6t(1 t)dt = [3t2 2t3 ]x = 3x2 2x3 , 0 x 1 0 and F (x) = 0 otherwi e. (b) 0.396 27.14 (a) The den ity function i given by f (x) = 6! 3 x (1 x)2 = 60x3 (1 x)2 , x > 0 2!3!
487 and 0 otherwi e. (b) For 0 < x < 1, we have F (x) = 60 x4 2 5 1 6 x + x 4 5 6 For x < 0, F (x) = 0 and for x 1, F (x) = 1 27.15 0.472 27.16 (a) For 0 < x < 1 we have f (x) = 43.649x1.0952 (1 x)4.6648 and 0 otherwi e. (b) 0.083 27.17 (a) 1 2 (b) 0.6875 27.18 We have E(X n ) = 27.19 We have FY (y) = P (Y y) = P (X > 1 y ) = 1 FX (1 y) Di erentiate with re pect to y to obtain 1 fY (y) = FX (1 y) = y b1 (1 y)a1 b That i , Y i a beta di tribution with parameter (b, a) 27.20 (a) 1 (b ) 0.2373046875 6 27.21 0.264 27.22 42 27.23 (a) The den ity function i f (x) = The expected value i E(X) = Section 28 1 3 1 B(a, b) 1 xa+n1 (1 x)b1 dx = 0 B(a + n, b) B(a, b) 2(1 x) 0 x 1 0 el ewhere (b) 0.91
488 28.1 fY (y) = 28.2 fY (y) = a 1 e 2 a 2 2(y+1) and fY (y) 9 ( yb )2 ANSWER KEYS = 0 otherwise 1 3
28.3 or 0 y 8 we have fY (y) = y 6 and 0 otherwise 28.4 or y 1 we have fY (y) = y 1 and 0 otherwise c 28.5 For y 0 we have fY (y) = m 2y e m and 0 otherwise m 28.6 For y > 0 we have fY (y) = ey and 0 otherwise 1 28.7 For 1 y 1 we have fY (y) = 1 2 and 0 otherwis 1 and 0 otherwise, E(Y ) = +1 28.8 ( ) For 0 < y < 1 we h ve fY (y) = (b) For y < 0 we h ve fY (y) = ey nd 0 otherwise, E(Y ) = 1 e 1 (c) For 1 < y < e we h ve fY (y) = y nd 0 otherwise, E(Y ) = 1 dy = e 1. 2 (d) For 0 < y < 1 we h ve fY ( y) = 2 2 , E(Y ) = 1y 1y 1 1 y 2y 28.9 999 28.10 For y > 4 fY (y) = 4y 2 and 0 otherwise 28.11 For 10, 000e0.04 < v < 10, 000e0.08 FV (v) = FR (g 1 (v)) = 25 ln and FV (v) = 0 for v 10, 000e0.04 a nd FV (v) = 1 for v 10, 000e0.08 5 1 y 4 1 y 28.12 For y > 0 we have fY (y) = 8 10 4 e( 10 ) and 0 otherwise 5 5 28.13 fR (r) = 2r2 for 5 < r < 4 and 0 otherwise 6 y 1 28.14 fY (y) = 2 fX 2 1 1 28.15 (a) 0.341 ( ) fY (y) = 2y2 ex 222 (ln y 1)2 28.16 The cdf is given by 0 a0 a 0<a<1 FY (a) = 1 a1 Thus, the ensity function f Y is fY (y)(a) = FY (a) = 1 1 a 2 2 v 10,000 0.04 0 0<a<1 otherwise Hence, Y is a beta ran om variable with parameters ( 1 , 1) 2 28.17 (a) fY (a) = a2 18 0 3 < a < 3 otherwise
489 (b) fZ (a) = 28.18 fY (y) = 1 1, y 1 2y 2 e2y 3 (3 2 a)2 2 < a < 4 0 otherwise 0<y1 28.19 fY (y) = e 28.20 (a) fY (y) = 8Y for 2 y 8 an 0 elsewhere. 50 9 (b) We have E(Y ) = 4 (c) 25 3 Section 29 29.1 (a) From the table we see that the sum of al l the entries is 1. (b) 0.25 (c) 0.55 ( ) pX (0) = 0.3, pX (1) = 0.5, pX (2) = 0 .125, pX (3) = 0.075 an 0 otherwise 29.2 0.25 29.3 (a) 0.08 (b) 0.36 (c) 0.86 ( ) 0.35 (e) 0.6 (f) 0.65 (g) 0.4 29.4 (a) 0.4 (b) 0.8 29.5 0.0512 29.6 (a) The c f o X an Y is x y FXY (x, y) = = fXY (u, v) u v y x 2 2 v2 u 2 ue u ve 0 x2 y2
0 =(1 e 2 )(1 e 2 ) an 0 otherwise. (b) The marginal p f for X is fX (x) = fXY (x, y) y = 0 xye x2 +y 2 2
y = xe 2
x2 an the marginal p f for Y is fY (y) = fXY (x, y) x = 0 xye x2 +y 2 2
x = ye 2
y2 for y > 0 an 0 otherwise 29.7 We have 1 1 fXY (x, y) x y =

x y a 1a
xa y 1a
=(2 + a a ) 2 1 =1

x y = 0 0 x y
490 ANSWER KEYS
so fXY (x, y) is not a ensity function. However one can easily turn it into a ensity function by multiplying f (x, y) by (2 + a a2 ) to obtain the ensity fun ction (2 + a a2 )xa y 1a 0 x, y 1 fXY (x, y) = 0 otherwise 29.8 0.625 29.9 0.708 29.10 0.576 29.11 0.488 3 1 y 29.12 fY (y) = y 15y x = 15y 2 (1 y 2 ), 0 < y < 1 an 0 otherwise 29.13 5.778 29.14 0.83 29.15 0.008 7 29.16 20 29.17 1 2e1 12 29. 18 25 29.19 3 8 29.20 (a) We have X\ Y 1 2 pY (y) (b) We have x < 1 or y < 1 0 0 .2 1 x < 2 an 1 y < 2 0.7 1 x < 2 an y 2 FXY (x, y) = 0.4 x 2 an 1 y < 2 an y 2 29.21 1 2 1 2 pX (x) 0.2 0.5 0.7 0.2 0.1 0.3 0.4 0.6 1 1 1 2 0 f (s, t) s t + 1 0 1 2 0 f (s, t) s t Section 30 30.1 (a) Yes (b) 0.5 (c) 1 ea 30.2 (a) The joint ensity over the regi on R must integrate to 1, so we have 1= (x,y)R c x y = cA(R).

491 1 (b) Note that A(R) = 4 so that fXY (x, y) = 4 = 1 1 . Hence, by Theorem 22 30. 2, X an Y are in epen ent with each istribute uniformly over (1, 1). 1 (c) P ( X 2 + Y 2 1) = x y = 4 x2 +y 2 1 4 30.3 (a) 0.48475 (b) We have 1 1 x fX (x) = x 6(1 y)dy = 6y 3y 2 = 3x2 6x + 3, 0 x 1 and 0 otherwise. Similarly, y fY (y) = 0 y2 6(1 y)dy = 6 y 2 y = 6y(1 y), 0 y 1 0 and 0 otherwise. (c) X and Y are de endent 30.4 (a) k = 4 (b) We have 1 fX (x) = 0 4xydy = 2x, 0 x 1 and 0 otherwise. Similarly, 1 fY (y) = 0 4xydx = 2y, 0 y 1 and 0 otherwise. (c) Since fXY (x, y) = fX (x)fY (y), X and Y are inde endent. 3 0.5 (a) k = 6 (b) We have 1 fX (x) = 0 6xy 2 dy = 2x, 0 x 1, 0 otherwise 1 and fY (y) = 0 6xy 2 dy = 3y 2 , 0 y 1, 0 otherwise (c) 0.15 (d) 0.875 (e) X and Y are inde endent 8 30.6 (a) k = 7 (b) Yes (c) 16 2 1 30.7 (a) We have 2

fX (x) = 0 3x2 + 2y 6x2 + 4 dy = , 0 x 2, 0 otherwise 24 24
492 and 2 ANSWER KEYS fY (y) = 0 3x2 + 2y 8 + 4y dx = , 0 y 2, 0 otherwise 24 24 (b) X and Y are de endent. (c) 0.340 30.8 (a) We have 3x fX (x) = x 4 4 8 3 dy = x, 0 x , 0 otherwise 9 3 9 2 and 4 y 9 4 (3 9 fY (y) = 0y 3 2 y) 3 y 3 2 0 otherwise (b) 2 (c) X and Y are de endent 3 30.9 0.469 30.10 0.191 30.11 0.4 30.12 0.19 30 .13 0.295 30.14 0.414 1 30.15 f (z) = e 2 z ez , z > 0, 0 otherwise 2 30.16 f (x) = (2x+1)2 , x > 0, 0 otherwise 3 30.17 5 30.18 Su ose that X and Y are inde end ent. Then P (X = 0 Y = 1) = P (X = 0) = 0.6 and P (X = 1 Y = 0) = 0.7. Since P ( X = 0) + P (X = 1) = 0.6 + 0.7 = 1, it follows that X and Y can not be independe nt. 1 30.19 1 = 4 and 2 = 0 30.20 0.172 Section 31
493 31.1 P (Z = 0) =P (X = 0)P (Y = 0) = (0.1)(0.25) = 0.025 P (Z = 1) =P (X = 1)P (Y = 0 ) + P (Y = 1)P (X = 0) =(0.2)(0.25) + (0.4)(0.1) = 0.09 P (Z = 2) =P (X = 1)P (Y = 1) + P (X = 2)P (Y = 0) + P (Y = 2)P (X = 0) =(0.2)(0.4) + (0.3)(0.25) + (0.3 5)(0.1) = 0.19 P (Z = 3) =P (X = 2)P (Y = 1) + P (Y = 2)P (X = 1) + P (X = 3)P ( Y = 0) =(0.3)(0.4) + (0.35)(0.2) + (0.4)(0.25) = 0.29 P (Z = 4) =P (X = 2)P (Y = 2) + P (X = 3)P (Y = 1) =(0.3)(0.35) + (0.4)(0.4) = 0.265 P (Z = 5) =P (X = 3)P (X = 2) = (0.4)(0.35) = 0.14 and 0 otherwise 31.2 pX+Y (k) = 30 0.2k 0.830k or 0 k 30 and 0 otherwise. k 31.3 pX+Y (n) = (n 1)p2 (1 p)n2 , n = 2, and pX+Y (a) = 0 otherwise. 31.4 pX+Y (3) =pX (0)pY (3) = 1 1 1 = 3 4 12 4 12 4 pX+Y (5) =pX (1)pY (4) + pX (2)pY (3) = 12 3 pX+Y (6) =pX (2)pY (4) = 12 pX+Y (4) =pX (0)pY (4) + pX (1)pY (3) =
and 0 otherwise. 1 31.5 64 31.6 0.03368 31.7 P (X + Y = 2) = e p(1 p) + e p (b) P (Y > X) = ep
494 31.8 pX+Y (0) =pX (0)pY (0) = 1 1 1 = 2 2 4 ANSWER KEYS 1 1 1 1 1 + = 2 4 4 2 4 5 pX+Y (2) =pX (0)pY (2) + pX (2)pY (0) + pX (1)pY (1) = 16 1 pX+Y (3) =pX (1)pY (2) + pX (2)pY (1) = 8 1 pX+Y (4) =pX (2)pY (2) = 16 pX +Y (1) =pX (0)pY (1) + pX (1)pY (0) = 31.9 pX+Y (1) =pX (0)pY (1) + pX (1)pY (0) = 1 6 5 18 6 18 3 18 pX+Y (2) =pX (0)pY (2) + pX (2)pY (0) + pX (1)pY (1) = pX+Y (3) =pX (0)pY (3) + pX (1)pY (2) + pX (2)pY (1) + pX (3)pY (0) = pX+Y (4) =pX (0)pY (4) + pX (1)pY (3) + pX (2)pY (2) + pX (3)pY (1) + pX (4)pY ( 0) = pX+Y (5) =pX (0)pY (5) + pX (1)pY (4) + pX (2)pY (3) + pX (3)pY (2) 1 +pX ( 4)pY (1) + pX (4)pY (1) = 18 31.10 We have a pX+Y (a) = n=0 p(1p)n p(1p)an = (a+1)p2 (1p)a = C(a+1, a)p2 (1p)a
Thus, X + Y is a negative binomial with parameters (2, p). 31.11 9e8 10 31.12 e10 ( 10) 10! 31.13 2ea (1 ea ) 0a fX+Y (a) = 0 otherwise 31.14 1 ea 0a1 a e ( therwise
495 31.15 fX+2Y (a) = fX (a 2y)fY (y)dy 3 3 31.16 If 0 a 1 then fX+Y (a) = 2a 2 2 + a6 . If 1 a 2 then 7 fX+Y (a) = 6 a . If 2 a 3 then fX+Y (a) = 9 9 a + 3 a2 1 a3 . If a > 3 2 2 2 2 6 then fX+Y (a) = 0. 2 31.17 If 0 a 1 then fX+Y (a) = 3 a3 . If 1 < a < 2 then fX+Y (a) = 2 a3 + 4a 8 . If a 2 then fX+Y (a) = 0. 3 3 31. 18 fX+Y (a) = ea e . 31.19 fZ+T2 ( ) = e 2 e , > 0 nd 0 otherwise. 31. then fX+Y ( ) = 1 . If 4 6, then fX+Y ( ) = 3 4 2 2 4 nd fX+Y ( ) = 0 ot herwise. 2 2 31.21 If 0 < then fX+Y ( ) = 8 . If 2 < < 4 then fX+Y ( ) = 8 + 2 nd 0 otherwise. 2 2 2 31.22 fX+Y ( ) = 12 2 e( (1 +2 )) /[2(1 +2 )] . 2(1 +2 ) 31.23 31.24 fT (t) = 31.25 1 2e1 1 8 t t e d 0 = tet . Section 32 32.1 pX Y (0 1) = 0.25 and pX Y (1 1) = 0.75 32.2 (a) For 1 j 5 and i = 1, , j then we have pXY (j, i) = (b) pX Y (j i) = . ( ) (c) X and Y are depen dent 32.3 1 5j 5 1 k=i 5k 1 5 1 j P (X = 3, Y = 4) 0.10 2 = = P (Y = 4) 0.35 7 P (X = 4, Y = 4) 0.15 3 P (X = 4 Y = 4) = = = P (Y = 4) 0.35 7 P (X = 5, Y = 4) 0.10 2 P (X = 5 Y = 4) = = = P (Y = 4) 0.35 7 P (X = 3 Y = 4) =
496 32.4 P (X Y = 1) = P (Y P (X = 3 Y = pY X (4 y) 2 1 3 2 5 2 7 2 ANSWER KEYS
= 0 Y = 1) = P (X = 0, Y = 1) P (Y = 1) P (X = 1, Y = 1) P (X = 1 = 1) P (X = 2, Y = 1) P (X = 2 Y = 1) = P (Y = 1) P (X = 3, Y = 1) 1) = P (Y = 1) y 1 pY X (1 y)) 1 pY X (2 y) 2 3 pY X (3 y) 2 5 7 pY X (5 y) 2 9 2 pY X (6 y) 11 2 0 9 2 11
1/16 6/16 3/16 = 6/16 2/16 = 6/16 0/16 = 6/16 = 5 0 0 0 0 1 9 2 11 1 6 1 = 2 1 = 3 = =0 32.5 3 0 0 1 5 2 7 2 9 2 11 4 0 0 0 1 7 2 9 2 11 6 0 0 0 0 0 1 11 X and Y are dependent. 32.6 (a) pY (y) = C(n, y)py (1 p)ny . Thus, Y is a binomia l distribution with parameters n and p. (b) pX Y (x y) = pXY (x, y) pY (y) n!y x (pe1 )y (1p)ny y!(ny)!x! = C(n, y)py (1 p)ny y x ey , x = 0, 1, 2, x! =0, otherwise = Thus, X Y = y is a Poisson distribution with parameter y. X and Y are not independent. 32.7 x=0 1/11 pXY (x, 0) 4/11 x=1 pX Y (x 0) = = x=2 6/11 pY (0) 0 otherwise
497 x=0 3/7 pXY (x, 1) 3/7 x=1 pX Y (x 1) = = x=2 1/7 pY (1) 0 otherwise The ditional probability distribution for Y given X = x is 1/4 y=0 pXY (0, y) 3/4 y= 1 pY X (y 0) = = pX (0) 0 otherwise 4/7 y=0 pXY (1, y) 3/7 y=1 pY X (y 1) = = pX (1) 0 otherwise 6/7 y=0 pXY (2, y) 1/7 y=1 = pY X (y 2) = pX (2) 0 otherwise 2 32.8 (a) 2N11 (b) pX (x) = 2N 1 (c) pY X (y x) = 2x (1 2x )y 1 n 32.9 P (X = k X + Y = n) = C(n, k) 2 32.10 (a) x P (X = 0, Y = 0) = 48 47 52 51 48 4 P (X = 1, Y = 0) = 52 51 4 48 P (X = 1, Y = 1) = 52 51 4 3 P (X = 2, Y = 1) = 52 51 188 221 16 = 221 16 = 221 1 = 221 = and 0 otherwise. 204 (b) P 12 and 13 1 P (Y = 1) = P e. 16 1 1 (c) pX Y (1 1) = ise 17 Section 33 33.1 For (Y = (X = 13 y 0) = P (X = 0, Y = 0) + P 1, Y = 1) + P (X = 2, Y = 221 = 16 and pX Y (2 1) = x 1, 0 y 1 we have (X 2) 13 fX = = Y 1, Y = 0) = 221 = 13 and 0 otherwis 221 = 17 and 0 otherw (x y) = 3 x2 2 1 y3
498 If y = 0.5 then 12 2 x , 0.5 x 1. 7 The graph of fX Y (x 0.5) is given bel ow fX Y (x 0.5) = ANSWER KEYS 33.2 fX Y (x y) = 33.3 fY X (x y) = 0y<x1 33.4 fX Y (x y) = (y + 1)2 xex(y+1) , x 0 an fY X (x y) = xexy , y 0 2 33.5 (a) For 0 < y < x we have fXY (x, y) = y2 ex (b) fX Y (x y) = e(xy) , 0 < y < x 33.6 (a) fX Y (x y) = 6x(1 x), 0 < x < 1. X and Y are independen t. (b) 0.25 1 xy+1 33.7 fX Y (x y) = 3 3 y (b) 11 24 2 33.8 0.25 33.9 8 9 33.10 0. 4167 33.11 7 8 33.12 0.1222 1 1 33.13 fX (x) = x2 y dy = 2(1 x), 0 < x < 1 and 0 otherwise 1 33.14 mean= 1 and Var(Y ) = 18 3 1 33.15 1y for 0 < y < x < 1 Section 34 34.1 fZW (z, w) = fXY ( zdbw , awcz ) adbc adbc adbc 2 y1 = y2 e , y2 > 2x , y2 3y 2 , x3 0x<y1
34.2 fY1 Y2 (y1 , y2 ) 1, y1 n y2 34.3 fR (r, ) = r XY (r cos , r sin ), r > 0, 4.4 z [fXY (z( 1 + w2 )1 , wz( 1 + w2 )1 ) fZW (z, w) = 1 + w2 +fXY (z( 1 + w2 ) z( 1 + w2 )1 )]
499 (u) v (1v) (+) X 34.5 fU V (u, v) = e (+) . Hence X + Y and X+Y ()() are indep + Y having a gamma distri ution with parameters X ( + , ) and X+Y having a beta dis tribution with parameters (, ). 34.6 ey1 y1 y1 0, 0 < y2 < 1 fY1 Y2 (y1 , y2 ) = 0 otherwise 1 34.7 fY1 Y2 (y1 , y2 ) = 1 e( 7 y1 + 7 y2 ) /2 1 e( 7 y1 7 y2 ) /8 7 2 4.8 We have u = g1 (x, y) = 2y cos x and v = g2 (x, y) = 2y sin x. The Jacobian of the transformation is 2y sin x cos x 2y J= = 1 sin x 2y cos x 2y 3 1 2 1 2 2 u +1 1 1
Also, y= Thus, 1 u2 +v2 e 2 . 2 This is the joint density of two inde endent stan dard normal random variables. 34.9 fX+Y (a) = fXY (a y, y)dy If X and Y are inde endent then fX+Y (a) = fX (a y)fY (y)dy 34.10 fY X (a) = fXY (y a, y)dy. If are inde endent then fY X (a) = fX (y a)fY (y)dy = fX (y)fY (a + y)dy 1 34. u) = v fXY (v, u )dv. If X and Y are independent then fU (u) = v 1 fX (v)fY ( u )dv v v 1 34.12 fU (u) = (u+1)2 for u > 0 and 0 elsewhere. fU V (u, v) = fXY ( x(u, v), y(u, v)) J 1 = Section 35 7 35.1 (m+1)(m1) 35.2 E(XY ) = 12 3m 1 35.3 E( X Y ) = 3 7 5 35.4 E(X 2 Y ) = 36 and E(X 2 + Y 2 ) = 6 . 35.5 0 35.6 33 35.7 L 3 35.8 1 u2 + v 2 2
500 35.9 30 19 35.10 (a) 0.9 (b) 4.9 (c) 4.2 35.11 (a) 14 (b) 45 35.12 5.725 35. 13 2 L2 3 35.14 27 ANSWER KEYS Section 36 36.1 2 2 36.2 coveraince i 0.123 and correlation i 0.33 1 36.3 covaria nce i 12 and correlation i 22 36.4 (a) fXY (x, y) = 5, 1 < x < 1, x2 < y < x2 + 0.1 and 0 otherwi e. (b) covariance i 0 and correlation i 0 36.5 We have 2 1 cos d = 0 E(X) = 2 0 E(Y ) = E(XY ) = 1 2 2 sin d = 0 0 2 1 2 cos sin d = 0 0 Thus X and Y are uncorrelated, but they are clearly not independent, since they are both unctions o 36.6 (a) (X1 + X2 , X2 + X3 ) = 0.5 (b) (X1 + X2 , X3 + X4 ) = 0 n 36.7 36 36.8 We have 1 1 xdx = 0 E(X) = 2 1 E(XY ) = E(X 3 ) = 1 2 1 x3 dx = 0 1 Thus, (X, Y ) = Cov(X, Y ) = 0 3 36.9 Cov(X, Y ) = 160 and (X, Y ) = 0.397 36.10 2 4 36.11 19,300 36.12 11 36.13 200 36.14 0
501 36.15 36.16 36.17 36.18 36.19 0.04 6 8.8 0.2743 (a) fXY (x, y) = fX (x)fY (y ) = (b) 0 a 2 0 a0 0<a1 1<a2 2<a3 a>3 0 < x < 1, 0 < y < 2 0 othe 1 2 fZ (a) = 1 2 3a 2 1 (c) E(X) = 0.5, Va (X) = 12 , E(Y ) = 1, Va (Y ) = 1 3 5 (d) E(Z) = 1.5 and Va (Z) = 12 36.20 (a) fZ (z) = dFz (z) = z and 0 othe wise. dz 2 (b) fX (x) = 1 x , 0 < x < 2 and 0 othe wise; fY (y) = 1 y , 0 < y < 2 2 2 and 0 othe wise. 2 (c) E(X) = 3 and Va (X) = 2 9 (d) Cov(X, Y ) = 1 9 36.21 0.15 1 36.22 6 5 36.23 12 3 6.24 1.04 36.25 2 36.26 E(W ) = 4 and Var(W ) = 67 36.27 1 5 36.28 2 n2 36.29 n+2 36.30 (a) We have X\Y 0 1 2 PY (y) 0 0.25 0.12 0.03 0.4 1 0.08 0.20 0.07 0.35 2 0.05 0.10 0.10 0.25
X (x) 0.38 0.42 0.2 1
502 (b) E(X) = 0.82 and E(Y ) = 0.85. (c) 0.243 (d) $147.75 36.31 mean is 140 an d standard deviation is 29.4 Section 37 2 37.1 E(X Y ) = 3 y and E(Y X) = 1x3 1x2 ANSWER KEYS 2 3 1 37.2 E(X) = 1 , E(X 2 ) = 1 , V ar(X) = 12 , E(Y X) = 3 x, V ar(Y X) = 2 3 4 3 2 1 3 19 x , E[V ar(Y X)] = 80 , V ar[E(Y X)] = 64 , Var(Y ) = 320 80 37.3 + 37.4 0.75 37.5 2.3 6 37.6 2 1x4 3 1x 37.7 A rst way for nding E(Y ) is 1 E(Y ) = 0 7 5 y y 2 dy = 2 1 0 7 7 7 y y 2 dy = . 2 9 For the second way, we use the doub e expectation resu t 1 1 E(Y ) = E(E(Y X)) = 1 E(Y X)fX (x)dx = 1 2 3 1 x6 1 x4 7 21 2 x (1x6 ) = 8 9 37.8 E(X) = 15 and E(Y ) = 5 37.9 1.4 37.10 15 5 37.11 E(X Y = 1) = 2, E(X Y = 2) = 3 , E(X Y = 3) = X and Y are dependent. 37.12 0.20 1 37.13 12 37.14 0.9856 37 .15 13 2 37.16 (1x) 12 37.17 Mean is and variance is 2 ( + 2 ) + 2 Section 38 2 E(X) = n+1 and Var(X) = n 12 2 1 38.2 E(X) = p and Var(X) = 1p p2 38.3 The momen t generating function is x pXY (x,1) pY (3) = 12 . 5 MX (t) = n=1 etn 6 2 n2 .
503 By the ratio test we have lim 6 et(n+1) 2 (n+1)2 n etn 26n2 = lim et n n2 = et > 1 (n + 1)2 and so the summation diverges whenever t > 0. Hence there does not exist a neigh borhood about 0 in which the mgf is nite. 38.4 E(X) = nd V r(X) = 2 38.5 Let Y = X1 + X2 + + Xn where each Xi is an exponential random variable with parameter . T hen n n MY (t) = k=1 MXk (t) = k=1 t = t n , t < .
Since this is the mgf of a gamma random variab e with parameters n and we can co nc ude that Y is a gamma random variab e with parameters n and . 1 t=0 38.6 MX (t ) = {t R : MX (t) < } = {0} otherwise 38.7 MY (t) = E(etY ) = e2t 3t , 3t < 3 3 his is a binomia random variab e with p = 4 and n = 15 38.9 Y has the same dist ribution as 3X 2 where X is a binomia distribution with n = 15 and p = 3 . 4 38. 10 38.11 38.12 38.13 38.14 38.15 E(t1 W + t2 Z) = e 2 e 2 = et1 +t2 5,000 10,560 8 M (t) = E(ety ) = 19 + 27 et 27 0.84 pet (a) MXi (t) = 1(1p)et , t < n (1 p) t (t1 +t2 )2 (t1 t2 )2 2 2
pe (b) MX (t) = 1(1p)et , t < n (1 p). (c) Because X1 , X2 , , Xn are independ then n n MY (t) = k=1 MXi (t) = pet 1 (1 p)et k=1 = pet 1 (1 p)et n n
504 t ANSWER KEYS n pe is the moment generating function of a negative binoBecause 1(1p)et mial random variable with parameters (n, p) then X1 + X2 + + Xn is a negative binomial rand om variable with the same pmf 38.16 0.6915 38.17 2 t 2 38.18 MX (t) = e (66t+3t )6 t3 38.19 38 1 38.20 2e 38.21 38.22 38.23 38.24 38.25 38.26 38.27 38.28 38.29 38.30 e2 4t k2 e 0.4 15 16 (0.7 + 0.3et )9 0.3346 41.9 (e 1 1)(et2 1) t1 t2 2 9 13t2 +4t
Section 39 39.1 Clearly E(X) = 2 + 2 = 0, E(X 2 ) = 2 and V ar(X) = 2 . Thus, 2 P ( X 0 ) = 1 = 2 = 1 39.2 100 39.3 0.4444 3 39.4 P (X 104 ) 104 = 0.1 10 20 39. 5 P (0 < X < 40) = P ( X 20 < 20) = 1 P ( X 20 20) > 1 202 = 39.6 P (X1 + X2 + + X20 > 15) 1 39.7 P ( X 75 10) = 1 P ( X 75 10) 1 39.8 Using Mark ty we n P (X ) = P (e tX t 19 20 25 100 = 3 4 E(etX ) e ) et
50 39.9 P (X > 75) = P (X 76) 76 0.658 8 39.10P (0.475 X 0.525) = P ( X 0.5 5) 1 625106 = 0.99 62510 39.11 By Markovs inequality P (X 2) E(X) = 2 = 1 2
505 1 1 39.12 100 ; P ( X 100 30) 302 = 180 an P ( X 100 < 30) 1 180 = 121 179 Therefore, the probability that the factorys pro uction will be between 180 70 an 130 in a ay is not smaller than 179 180 84 39.13 P (100 X 140) = P ( X 120 2 0) 1 202 0.79 2 39.14 We have E(X) = 0 x(2x) x = 3 < an E(X 2 ) = 0 x2 (2x) x = 1 1 7 < so that Var(X) = 18 4 = 18 < . Thus, by the Weak Law of 18 9 Large Numbe rs we know that X converges in probability to E(X) = 2 3 39.15 (a) 1 1 2
FYn (x) =P (Yn x) = 1 P (Yn > x) =1 P (X1 > x, X2 > x, , Xn > x) =1 P (X1 > x (X2 > x) P (Xn > x) =1 (1 x)n for 0 < x < 1. Also, FYn (x) = 0 for x 0 and FYn x) = 1 for x 1. (b) Let > 0 be given. Then P ( Yn 0 ) = P (Yn ) = 1 1 n 1 (1 < < 1 Consi ering the non trivial case 0 < < 1 we n n
lim p( Yn 0 ) = lim [1 (1 )n ] = 1 lim 1 0 = 1. n n Hence, Yn 0 in probability. Section 40 40.1 0.2119 40.2 0.9876 40.3 0.0094 40.4 0.692 40.5 0.1367 40.6 0.383 40.7 0.0088 40.8 0 40.9 23 40.10 0.0162 40.11 6,342 ,637.5
506 ANSWER KEYS
40.12 0.8185 40.13 16 40.14 0.8413 40.15 0.1587 40.16 0.9887 40.17 0.0475 40.18 (a) X is approximated by a normal distribution with mean 100 and variance 400 = 4. (b) 0.9544. 100 40.19 (a) 0.79 (b) 0.9709 Section 41 41.1 (a) 0.167 (b) 0.583 3 (c) 0.467 (d) 0.318 41.2 For t > 0 we have P (X a) eta (pet + 1 p)n an for t < 0 we have P (X a) eta (pet + 1 p)n 41.3 0.692 41.4 0.0625 41.5 For t > 0 we have P (X n) ent e(t1) and for t < 0 we have P (X n) ent e(t1) 41.6 (a) 0.769 (b) e26t e20(t1) (c) 0.357 (d) 0.1093 41.7 Fo ow from Jensens inequa ity a a 41.8 (a) a1 (b) a+1 1 2 (c) We have g (x) = x2 and g (x) = x3 . Since g (x) > 0 for a x in (0, ) we conc ude that g(x) is convex there. (d) We have a1 a2 1 1 = = E(X) a a(a + 1) and 1 a a2 E( ) = = . X a+1 a(a + 1) a a 1 1 1 Since a2 a2 1, we have a(a+1) a(a+1) . That is, E( X ) E(X) ), which ve ri es Jensens inequa ity in this case. 1 41.9 Let X be a random variab e such that p(X = xi ) = n for 1 i n. 2 Let g(x) = n x . By Jensens inequa ity we have for X > 0 2 2 E[ n (X 2 )] n [E(X 2 )]. That is E[ n (X 2 )] n [E(X 2 )].
507 But 1 E[ n (X )] = n 2 n n x2 = i i=1 1 n (x1 x2 xn )2 . n and ln [E(X 2 )] = ln It follows that x2 + x2 + + x2 1 2 n n x2 + x2 + + x2 n 2 1 n x2 + x2 + + x2 1 2 n n ln (x1 x2 xn ) n ln or (x1 x2 xn ) n Section 43 1 x+1 43.1 0 x 1 0 x 2 2 2 0 f (x, y)dydx or 1 2 0 1 2y f (x, y)dxdy 43.3 43.4 1 1 2 1 2 1 1 2 f (x, y)dydx 43.5 43.6 43.7 1 2e 1 43.8 6 4 43.9 L 3 1 1 43.10 0.5 0.5 f (x, y)dydx 7 43.11 8 3 43.12 8 1 1 1 f (x, y)dydx 1 f (x, y)dydx + 1 0 0 2 2 30 50x f (x, y)dydx 20 20 1 x+1 x+1 f (x, y)dydx + 1 x1 f (x, y)dydx 0 0 1
508 ANSWER KEYS
Bibliography [1] Sheldon Ross, A rst course in Probability, 6th Edition (2002), Prentice Hall. [2] SOA/CAS, Exam P Sample Questions 509

Marcel Finan S Exam 1P Manual

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Marcel Finan S Exam 1P Manual

Uploaded by

Copyright:

Available Formats

A Probability Course for the Actuaries A Preparation for Exam P/1 Marcel B.

Finan Arkansas Tech University c All Rights Reserved Preliminary Draft

2 In memory of my parents August 1, 2008 January 7, 2009

18 BASIC OPERATIONS ON SETS where the or is inclusive.(See Figure 2.2(a))

2 SET OPERATIONS (g) A (Ac B c ) 19

20 BASIC OPERATIONS ON SETS

22 BASIC OPERATIONS ON SETS

24 BASIC OPERATIONS ON SETS

26 BASIC OPERATIONS ON SETS

28 BASIC OPERATIONS ON SETS

30 BASIC OPERATIONS ON SETS

32 COUNTING AND COMBINATORICS

3 THE FUNDAMENTAL PRINCIPLE OF COUNTING 33

4 PERMUTATIONS AND COMBINATIONS 37

5 PERMUTATIONS AND COMBINATIONS WITH INDISTINGUISHABLE OBJECTS47

48 COUNTING AND COMBINATORICS

52 COUNTING AND COMBINATORICS

56 COUNTING AND COMBINATORICS

6 BASIC DEFINITIONS AND AXIOMS OF PROBABILITY 61

68 PROBABILITY: DEFINITIONS AND PROPERTIES

OF PROBABILITY Proof. We have P (B) + P (C) P (B C) P B) + P (A C) P ((A B) ((A B) (A C)) = P (A) +

P (A B C) ((A B) (A (A C))] = P P (B) + P (C)

= P (A) + P (B C) P (A (B C)) = P (A) + P (B) + P (C) (A) + P (B) + P (C) P (B C) P (A B) P (A C) P (B

9 CONDITIONAL PROBABILITIES Proof. We have P (A1 A2 An ) =P ((A1 A2 An1 ) An ) =P (An A1 A2 An 81

82 CONDITIONAL PROBABILITY AND INDEPENDENCE

98 CONDITIONAL PROBABILITY AND INDEPENDENCE

100 CONDITIONAL PROBABILITY AND INDEPENDENCE

102 (b) Let CONDITIONAL PROBABILITY AND INDEPENDENCE A1 A2 A3 A4

here I{1,2,3,4} (x) =

124 DISCRETE RANDOM VARIABLES

126 and the average value of X is DISCRETE RANDOM VARIABLES

18 BINOMIAL AND MULTINOMIAL RANDOM VARIABLES 149

150 DISCRETE RANDOM VARIABLES

160 (b) DISCRETE RANDOM VARIABLES

162 DISCRETE RANDOM VARIABLES To nd the variance, we rst compute E(X 2 ).

364 365 100 0.2399.

176 DISCRETE RANDOM VARIABLES

180 and the variance is Var(X) = DISCRETE RANDOM VARIABLES r(1 p) = 90 p2

21 PROPERTIES OF THE CUMULATIVE DISTRIBUTION FUNCTION189

n 1. De ne the events F1 =E1 c Fn =En E

196 DISCRETE RANDOM VARIABLES Figure 21.2

198 DISCRETE RANDOM VARIABLES

21 PROPERTIES OF THE CUMULATIVE DISTRIBUTION FUNCTION199

200 DISCRETE RANDOM VARIABLES

202 DISCRETE RANDOM VARIABLES

204 CONTINUOUS RANDOM VARIABLES

Figure 22.1 Now, if we let

= b in the previous formul we

< X < b) = P ( cumu ative distribu ned as fo ow variab e X as we can write

22 DISTRIBUTION FUNCTIONS 205

robability P (1 X 1) is calculated as follo

dx = 69 128 (c) For 2 x 0 we have x F (x) = 2 15 t + 64 64 dt = x2 15 7 + x+ 128 64 16

Given r ndom loss .5 is equ l to 0.64 endent, identic lly ) = 3x2 0 x 1 0

0 E(Y ) = 0 P (Y > y)dy P (Y < y)dy.

222 CONTINUOUS RANDOM VARIABLES

Ex mple 23.6 A m nuf cturers ction f (x) = 2.5(0.6)2.5 x3.5

nnu l losses follow

distribution with density fun

Solution. Note for x

ny of the r ndom v ri bles the cdf is given by

x 0 dx = 1 x 2 x 0 x2 = x0 x2 0 2 provided > 2. Hence, V r(X) = x2 2 x2 x2 0 0 0 = 2 ( 1)2 ( 2)( 1)2

Percentiles nd Interqu rtile R nge We de ne the 100pth percentile of to be the qu ntity th t