Chapter 8

Problems
1.

P{0 X 40} = 1 P{X 20 > 20} 1 20/400 = 19/20

2.

(a) P{X 85} E[X]/85 = 15/17

(b) P{65 X 85) = 1 P{X 75 > 10} 1 25/100

(c) P

3.

i =1

25
X i / n 75 > 5
so need n = 10
25n

Let Z be a standard normal random variable. Then,

4.

i =1

/ n 75 > 5 P{Z >

n } .1 when n = 3

20

(a) P X i > 15 20 / 15
i =1

20

20

(b) P X i > 15 = P X i > 15.5

i =1

i =1

15.5 20
P Z >

20

= P{Z > 1.006}

.8428

5.

50

Letting Xi denote the ith roundoff error it follows that E X i = 0,

i =1
50

Var X i = 50 Var(X1) = 50/12, where the last equality uses that .5 + X is uniform (0, 1)
i =1
and so Var(X) = Var(.5 + X) = 1/12. Hence,

{ X

> 3 P{N(0, 1) > 3(12/50)1/2} by the central limit theorem

= 2P{N(0, 1) > 1.47 = .1416

6.

If Xi is the outcome of the ith roll then E[Xi] = 7/2 Var(Xi) = 35/12 and so
79

79

P X i 300 = P X i 300.5
i =1

i =1

300.5 79(7 / 2)
P N (0,1)
= P{N (0,1) 1.58} = .9429
(79 35 / 12)1/2

141
Copyright 2014 Pearson Education, Inc.

142

7.

Chapter 8
100

525 500

P X i > 525 P N (0,1) >

= P{N (0,1) > .5} = .3085
(100 25)
i =1

where the above uses that an exponential with mean 5 has variance 25.
8.

If we let Xi denote the life of bulb i and let Ri be the time to replace bulb i then the desired
99
100

probability is P X i + Ri 550 . Since

X i + Ri has mean 100 5 + 99 .25 =
i =1
i =1

524.75 and variance 2500 + 99/48 = 2502 it follows that the desired probability is
approximately equal to P{N(0, 1) [550 524.75]/(2502)1/2} = P{N(0, 1) .505} = .693
It should be noted that the above used that

Var(Ri) = Var Unif [0,1] = 1/48

2

9.

Use the fact that a gamma (n, 1) random variable is the sum of n independent exponentials
with rate 1 and thus has mean and variance equal to n, to obtain:

{
}
P { N (0,1) > .01 n }
= 2 P {N (0,1) > .01 n }

X n

> .01 = P X n / n > .01 n

P
n

Now P{N(0, 1) > 2.58} = .005 and so n = (258)2.

10.

If Wn is the total weight of n cars and A is the amount of weight that the bridge can withstand
then Wn A is normal with mean 3n 400 and variance .09n + 1600. Hence, the probability
of structural damage is

P{Wn A 0} P Z (400 3n ) / .09n + 1600

Since P{Z 1.28} = .1 the probability of damage will exceed .1 when n is such that
400 3n 1.28 .09n + 1600
The above will be satisfied whenever n 117.
12.

Let Li denote the life of component i.

100
1
E Li = 1000 + 50(101) = 1505
10
i =1
2
100 100
i
1 100 2
2
10
(100)
(100)(101)
i
+
=
+
+
Var Li =

10
100 i =1
i =1 i=1

Now apply the central limit theorem to approximate.

Copyright 2014 Pearson Education, Inc.

Chapter 8

13.

143

X 74

(a) P{ X > 80} = P

> 15 / 7 PPZ > 2.14} .0162
14 / 5

Y 74

(b) P{Y > 80} = P

> 24 / 7 P{Z > 3.43} .0003
14 / 8

(c) Using that SD (Y X ) = 196 / 64 + 196 / 25 3.30 we have

P{Y X > 2.2} = P{Y X } / 3.30 > 2.2 / 3.30}
P{Z > .67} .2514

(d) same as in (c)

14.

Suppose n components are in stock. The probability they will last for at least 2000 hours is
n

2000 100n
p = P X i 2000 P Z

30 n

i =1

where Z is a standard normal random variable. Since

.95 = P{Z 1.64} it follows that p .95 if
2000 100n
1.64
30 n

or, equivalently,
(2000 100n)/ n 49.2
and this will be the case if n 23.
15.

10,000

P
X i > 2,700,000 P{Z (2,700,000 2,400,000)/(800 100)} = P{Z 3.75} 0
i =1

16.

(a) Number Aj's jobs, let X i be the time it takes to do job i, and let X A =

20
i =1

Xi

be the time that it takes AJ to finish all 20 jobs. Because

E [ X A ] = 20(50) = 1000,

Var ( X A ) = 20(100) = 2000

the central limit theorem gives that

X 1000 900 1000

P{ X A 900} = P A

2000
2000

= P{Z 2.236}
= 1 (2.236) = .013

Copyright 2014 Pearson Education, Inc.

144

Chapter 8
(b) Similarly, if we let X M be the that it takes MJ to finish all of her 20 jobs, then by the
central limit theorem X M is approximately normal with mean and variance.
E [ X M ] = 20(52) = 1040,

Var( X A ) = 20(225) = 4500

Thus
X 1040 900 1040

P{ X M 900} = P M

4500
4500

= P{Z 2.087}
= 1 (2.087) = .018

(c) Because the sum of independent normal random variables is also normal, D = X M X A
is approximately normal with mean and variance
E [ D ] = 1040 1000 = 40 ,

Var( D ) = 4500 + 2000 = 6500

Hence,
D 40
40
P ( D > 0) = P

6500
6500
= P{Z .4961}
= (.4961) = .691

Thus even though AJ is more likely than not to finish earlier than MJ, MJ, has the better
chance to finish within 900 minutes
19.

Let Yi denote the additional number of fish that need to be caught to obtain a new type when
4i
.
there are at present i distinct types. Then Yi is geometric with parameter
4
3
4 4
25
E[Y] = E Yi = 1 + + + 4 =
3 2
3
i =0

3 4
130
Var[Y] = Var Yi = + 2 + 12 =
9
i=0 9

Hence,
25 25 1300 1

PY
>

3
3
9 10

and so we can take a =

25 1300
25 + 1300
, b=
.
3
3

Copyright 2014 Pearson Education, Inc.

Chapter 8

145

Also,
25
130
1

P Y
when a =
> a
=
2
3
10

130 + 9a

1170
.
3

25 + 1170
Hence P Y >
.1.
3

21.

g(x) = xn(n1) is convex. Hence, by Jensens Inequality

E[Y n/(n1)] E[Y])n/(n1) Now set Y = X n1 and so
E[X n] (E[X n1])n/(n1) or (E[X n])1/n (E[X n1])1/(n1)

22.

No

23.

(a) 20/26 .769

(b) 20/(20 + 36) = 5/14 .357
(d) p P{Z (25.5 20)/ 20 } P{Z 1.23} .1093
(e) p = .112184

24.

Answer is (a). Because X is approximately normal with mean 100 and standard deviation 10
there is about a 2 percent chance that it will be more than two standard deviations larger than
its mean.

Copyright 2014 Pearson Education, Inc.

146

Chapter 8

Theoretical Exercises
1.

This follows immediately from Chebyshevs inequality.

2.

P{D >} = P{X > }

3.

(a)

(b)

2
1
= 2 2
2 2

r

np
np(1 p )

= np / (1 p )

(c) answer = 1
(d)

1/ 2
= 3
1 / 12

(e) answer = 1
(d) answer = /
4.

For > 0, let > 0 be such that g(x) g(c) < whenever x c . Also, let B be such
that g(x) < B. Then,
E[g(Zn)] =

x c

g ( x )dFn ( x ) +

x c >

g ( x )dFn ( x )

( + g(c))P{Zn c } + BP{Zn c > }

In addition, the same equality yields that
E[g(Zn)] (g(c) )P{Zn c } BP{Zn c > }
Upon letting n , we obtain that
lim sup E[g(Zn)] g(c) +
lim inf E[g(Zn)] g(c)
The result now follows since is arbitrary.
5.

Use the notation of the hint. The weak law of large numbers yields that
lim P{ ( X 1 + ... + X n ) / n c > } = 0

Since X1 + + Xn is binomial with parameters n, x, we have

X + ... + X n
Ef 1
n

f (k / n) k x

(1 x ) n k

k =1

The result now follows from Exercise 4.

Copyright 2014 Pearson Education, Inc.

Chapter 8

6.

147

E[X] =

i =1
k

i = k +1

i P{X = i} + i P{X = i}
i P{X = k}
i =1

= P{X = k}k(k + 1)/2

k2

P{ X = k }
2
7.

Take logs and apply the central limit theorem

8.

It is the distribution of the sum of t independent exponentials each having rate .

9.

1/2

10.

Use the Chernoff bound: etiM(t) = e ( e 1) ti will obtain its minimal value when t is chosen
to satisfy
t

et = i, and this value of t is negative provided i < .

Hence, the Chernoff bound gives
P{X i} ei(/i)i
11.

etiM(t) = (pet + q)neti and differentiation shows that the value of t that minimizes it is such
that
iq
npet = i(pet + q) or et =
(n i ) p
Using this value of t, the Chernoff bound gives that
n

iq

+ q ( n i )i p i / (iq )i
P{X i}
ni

=
12.

( nq )n ( n i )i pi
i i qi ( n i ) n

1 = E[eX] eE[X] by Jensens inequality.

Hence, E[X] 0 and thus > 0.

14.

Let each X i be Poisson with mean .01.

Copyright 2014 Pearson Education, Inc.