You are on page 1of 8

Practice Problems for PLCY 2455: Week 1

DESCRIPTIVE STATISTICS

(1) True / False: The U.S. News and World Report Graduate School edition listed the distribution of
private sector earnings of recent Yale Law School graduates as having a 25th percentile of $160,000 and a
75th percentile of $160,000. The mean and the median must also be the same.

UNCERTAIN. The median must also be $160,000, but the mean could be anything above $120,000
depending on the incomes in the top and bottom quartiles.

(2) True/False: You work in a large team. You don't know what the annual bonus will be, but your boss
has informed you she'll use a "bell curve" to determine the distribution of bonuses among the staff. After
landing a big client, your boss enthusiastically tells everybody their "annual bonus just doubled". The
median bonus, mean bonus, and variance all just doubled as well.

FALSE. The median and mean have doubled, but the variance has increased by a factor of 22 = 4.

1
PROBABILITY

(1) You are thinking of moving to Baltimore, home to one of the world's premier institutions of
medical research and training: Johns Hopkins. Unfortunately, the city also has some of the
highest rates of gang activity and murders per capita in the nation. In fact, the murder rate in
Baltimore is seven times the national average. During your interviews, one of the faculty
members says - "Don't worry. The violence, especially murder, is overwhelmingly among people
who are gang related." If all of the murders occurred among the gang related, a demographic
estimated to be 12.5% of city's the population, what would their murder rate be relative to the
national average?

(1/.125) * 7 = 8 * 7 = 56 times the national average

(2) If you ip a fair coin 3 times, what is the probability of


(a) getting all tails? ---> .5^10
(b) getting no tails? ---> this is like getting all heads = .5^10
(c) getting at least one tail? ---> this is the complement of getting all heads = 1 - .5^10

(3) In a class where everyones native language is English, 70% of students speak Spanish, 45%
speak French as a second language and 20% speak both. Assume that there are no students who
speak another second language.

(a) What percent of the students do not speak a second language?


1 - P(do speak 2nd) = 1 - [P(Spanish) + P(French) - P(French and Spanish)] = 1 - [.70 + .45 -
.2] = .05

(b) Are speaking French and Spanish mutually exclusive?


No. 20% speak both

(c) Are the variables speaking French and speaking Spanish independent?
This is equivalent to asking does P(F | S) = P(F) ? P(F | S) = .2/.7 = .29; P(F) = .45. Since the
two probabilities aren't the same, the two variables are not independent.

2
(4) Tarsiers are an endangered species of primate living in Southeast Asia. When female tarsiers
give birth, the probability distribution for the number of babies is:

# of Babies Probability
1 0.25
2 0.5
3 0.25

Suppose that you have two tarsier mothers who are preparing to give birth. You may assume that the
number of babies born to each tarsier mother is independent of the other.

a) What is the chance that at least one out of the two tarsier mothers will have two or more babies?
The probability that a single tarsier mother has two or more babies is 0.75. Then there are two
ways to solve for the probability that at least one mother out of two has two or more babies:
Complements Method:
P(At Least One 2+) = 1 - P(Neither 2+) = 1 - (1 - 0.75)2 = 1 - 0.0625 = 0.9375
Addition Rule Method:
P(At Least One 2+) = P(Tarsier #1 2+) + P(Tarsier #2 2+) - P(Both Tarsiers 2+)
= 0.75 + 0.75 - 0.75*0.75 = 1.5 - 0.5625 = 0.9375

b) What is the probability that both of the tarsier babies give birth to two or more babies?
From above, we have already solved for this probability in the addition rule method.
P(Both Tarsiers 2+) = 0.5625

c) What is the probability that the two tarsier mothers give birth to exactly four babies, between
them?
There are three ways for the two mothers to have exactly 4 babies between them:
Way #1: Mother #1: 1 Baby, Mother #2: 3 Babies: Prob = 0.25*0.25 = 0.0625
Way #2: Mother #1: 2 Babies, Mother #2: 2 Babies: Prob = 0.5*0.5 = 0.25
Way #3: Mother #1: 3 Babies, Mother #2: 1 Baby: Prob = 0.25*0.25 = 0.0625
Adding up the total probability is 0.25 + 2*0.0625 = 0.375.

Suppose that the probability of having one or three babies increases, while the probability of having two
babies decreases. You may assume that, as these probabilities change, the probability of having one baby
stays equal to the probability of having three babies.

d) Will the variance of the number of babies per litter increase, decrease, or stay the same? Justify
your answer.

3
The variance of babies will increase. Intuitively, variance is a measure of the spread of a distribution
around the mean. In this case, the mean number of babies is always 2. As the probability of having
either 1 or 3 babies increases, the distribution becomes more spread out, and the variance increases.
Mathematically, suppose that the probability of having 1 baby (which equals the probability of having
three babies) is q. Then the probability of having two babies is 1 - 2q. Since the mean is always 2, the
variance is
= ()( ()2 = (1 2)2 + (1 2)(2 2)2 + (3 2)2 = 2
The variance therefore increases as one increases q.

(5) TRANSPORTING DIGNITARIES

Say that you are tasked with transporting a group of visiting dignitaries, by road, from Harare to an exotic
conference site in Hwange National Park. You have two options: (a) put everybody in one vehicle (a
Toyota LandCruiser), in which case not everybody can have seatbelts, or (b) split the group between two
vehicles (two Toyota Hilux Surfs), in which case everybody can have seatbelts.
The primary danger is a tire blow-out. Given this particular trip, each tire has a 1% probability of
suffering a blow-out. If a blow-out occurs, a major injury could result. If everybody is in one car (and thus
some dont have seatbelts), then the probability of a major injury given a blow-out is 70%; if instead
passengers had been split between two cars, then the probability of a major injury given a blow-out is
instead 40%.
To simplify things, assume that your only concern is the potential for a major injury associated with a tire
blowout. Assume also that helicopters are dispatched and the road trip ends as soon as any tire blows out;
thus, you do not need to worry about multiple blow-outs. (To be clear, though: each tires probability of
blowing out is independent, and blow-out events are not disjoint.)
Please answer the following:
a) How would you write each of the probabilities presented above in formal notation?
P(blowout in 1 tire) = 0.01
P(injury | blowout in a 1-vehicle trip) = 0.7
P(injury | blowout in a 2-vehicle trip) = 0.4

b) What is the probability of suffering at least one blow-out under option (a) (one vehicle)? What is
the probability under option (b) (two vehicles)? Please report your answers in formal notation and
show any formulas or tables you used in your work.

Option (a): One-Vehicle Trip


First, note that if using one vehicle there are four wheels. Thus, to calculate the probability of a
blowout in a one-vehicle trip:

4
P(blowout in a 1-vehicle trip) = 1 P(no blowout in all 4 tires)
= 1 P(no blowout in 1 tire)4
= 1 (1 P(blowout in 1 tire)) 4
= 1 (1 0.01) 4
= 1 (0.99)4
= 1 0.9606
= 0.0394

Option (b): Two-Vehicle Trip


First, note that if using two vehicles there are eight wheels. Thus, to calculate the probability of a
blowout in a two-vehicle trip:
P(blowout in a 2-vehicle trip) = 1 P(no blowout in all 8 tires)
= 1 P(no blowout in 1 tire)8
= 1 (1 P(blowout in 1 tire)) 8
= 1 (1 0.01) 8
= 1 (0.99)8
= 1 0.9227
= 0.0773

Common question: Why cant we calculate the probability of a blowout by modeling the
probabilities of each of the tires blowing out as disjoint events?

Answer: The tire blowouts are not disjoint events. It is possible that multiple tires could blowout!
Even though the trip to Hwange National Park would stop as soon as one tire blew out, we have
to account for the possibility that multiple tires would have blown out, otherwise we are not fully
capturing all possible outcomes that could occur under our probability model. One way to think
of this is that we arent actually calculating the probability of a tire blowing out, but instead we
are modeling the probability of an event occurring that would lead to a tire blowing out. From this
perspective it is clear that multiple events could occur that each would lead to a tire blowing out.
Alternatively, we could decide to model the probability that each of the tires doesnt blow out.
This is a much easier quantity to calculate in this problem because there is only one way that all
the tires dont blow out, whereas there are fifteen different ways for at least one tire out of four to
blow out, (e.g. front right only, front left only, rear right only, rear left only, front right and front
left, etc.). When we consider eight wheels, there are 255 different ways that at least one tire could
blow out, but again only 1 way that no tires blow out. Because of this fact, we can use the
complement rule to calculate the probability of a blowout.

5
If you ever are confused in a problem like this, here is a good thing to try. Imagine that we had 25
cars, such that there are 100 tires between the cars. Is the probability of a blowout occurring now
100%? No way! What if we had 30 cars (and therefore 120 tires)?

c) What is the probability of suffering at least one blow-out and major injuries under option (a) (one
vehicle)? What is the probability under option (b) (two vehicles)? Please report your answers in
formal notation and show any formulas or tables you used in your work.

Option (a): one Toyota LandCruiser


Calculate the probability of a blowout and an injury occurring if using one vehicle:
P(blowout in a 1-vehicle trip AND injury in a 1-vehicle trip)
= P(injury | blowout in a 1-vehicle trip) * P(blowout in a 1-vehicle trip)
= 0.7 * 0.0394
= 0.0276

Option (b): two Toyota Hilux Surfs


Calculate the probability of a blowout and an injury occurring if using two vehicles:
P(blowout in a 2-vehcile trip AND injury in a 2-vehicle trip)
= P(injury | blowout in a 2-vehicle trip) * P(blowout in a 2-vehicle trip)
= 0.4 * 0.0773
= 0.0309

d) Considering only safety, which option would you choose? Explain in one or two sentences.
First note that we assume that an injury can only occur if there is a blowout, or more formally:
P(blowout | injury) = 1.
Therefore, the probability of an injury and a blowout simultaneously occurring is simply the
probability of an injury occurring, which we can show formally:
P(injury and blowout) = P(blowout | injury) * P(injury)
= 1 * P(injury)
= P(injury)
So the probabilities that we calculated in part c) are actually unconditional marginal probabilities
of an injury occurring under the two options, or more formally,
P(injury in a 1-vehicle trip) = P(blowout AND injury in a 1-vehicle trip)

6
= 0.0276

P(injury in a 2-vehicle trip) = P(blowout AND injury in a 2-vehicle trip)


= 0.0309

Now that we are dealing with unconditional marginal probabilities, we can make a direct
comparison and note that the 1-vehicle trip with the LandCruiser results in a lower probability of
injury than the 2-vehicle trip with the Hilux Surfs.

(6) Bicycle Security

During the orientation at a policy school, the Security Staff presented the following slide about Bicycle
Security.

(a) Explain to your stats professor the probability reported in the fourth bullet of the slide above.

The slide essentially says, consider all bicycle thefts, and note that a cable lock is cut off in 99
percent of them. Thus, the probability reported in the slide is the conditional probability P(cable
lock | theft) = 0.99.

(b) In a paragraph, explain to a friend whether the probability in the fourth bullet is the probability
you think would be most useful to an incoming policy student deciding whether to purchase a

7
cable lock or a U-lock kryptonite lock for his or her bike. Assume your friend is intelligent but
not well versed in statistics.
P(cable lock | theft) is not, by itself, a useful probability for an incoming student. To make an
informed decision about which type of lock to buy, an incoming student would want to compare
two different conditional probabilities: P(theft | cable lock) to P(theft | U-lock). These two
probabilities would allow for a direct assessment of the probable consequences of the decision of
which type of lock to buy.
Because we dont know the rate at which students use different types of locks, we cannot actually
assess which type of lock is associated with more thefts. As an extreme example, suppose 99.9
percent of students used a cable lock, and only 0.1 percent used a U-lock. If theft were unrelated
to the choice of lock, we would expect that 99.9 percent of thefts would be associated with cable
locks and 0.1 percent of thefts would be associated with U-locks. If we actually observe that only
99.0 percent of thefts were associated with cable locks and 1.0 percent of thefts were associated
with U-locks, we would conclude that cable locks actually seem better!

Problems from the Book

See the book for solutions to these problems.

You might also like