ExamQuestionsUP PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 103

Probability Exam Questions

for
Understanding Probability
by Henk Tijms
1
September 1, 2013
This note gives a large number of exam problems for a rst course in prob-
ability. Fully worked-out solutions of these problems are also given, but of
course you should rst try to solve the problems on your own!
Chapter 7
7E-1 A bridge hand in which there is no card higher than a nine is called a
Yarborough. Specify an appropriate sample space and determine the prob-
ability of Yarborough when you are randomly dealt 13 cards out of a well-
shued deck of 52 cards.
7E-2 Five dots are placed at random on a 8 8 grid in such a way that no
cell contains more than one dot. Specify an appropriate sample space and
determine the probability that no row or column contains more than one dot.
7E-3 Five people are sitting at a table in a restaurant. Two of them order
coee and the other three order tea. The waiter forgot who ordered what and
puts the drinks in a random order for the ve persons. Specify an appropriate
sample space and determine the probability that each person gets the correct
drink
7E-4 A parking lot has 10 parking spaces arranged in a row. There are 7
cars parked. Assume that each car owner has picked at a random a parking
place among the spaces available. Specify an appropriate sample space and
determine the probability that the three empty places are adjacent to each
other.
1
The exam questions are about the basic probability material in the chapters 7-14 of
my textbook Understanding Probability, third edition, Cambridge University Press, 2012.
The simulation-based chapters 1-6 of the book present many probability applications from
everyday life to help the beginning student develop a feel for probabilities. Moreover, the
book gives an introduction to discrete-time Markov chains and continuous-time Markov
chains in the chapters 15 and 16.
1
7E-5 A group of ve people simultaneously enter an elevator at the ground
oor. There are 10 upper oors. The persons choose their exit oors inde-
pendently of each other. Specify an appropriate sample space and determine
the probability that they are all going to dierent oors when each person
randomly chooses one of the 10 oors as the exit oor. How does the answer
change when each person chooses with probability
1
2
the 10th oor as the
exit oor and the other oors remain equally likely as the exit oor with a
probability of
1
18
each.
7E-6 Three friends and seven other people are randomly seated in a row.
Specify an appropriate sample space to answer the following two questions.
(a) What is the probability that the three friends will sit next to each other?
(b) What is the probability that exactly two of the three friends will sit next
to each other?
7E-7 You and two of your friends are in a group of 10 people. The group is
randomly split up into two groups of 5 people each. Specify an appropriate
sample space and determine the probability that you and your two friends
are in the same group.
7E-8 You are dealt a hand of four cards from a well-shued deck of 52 cards.
Specify an appropriate sample space and determine the probability that you
receive the four cards J, Q, K, A in any order, with suit irrelevant.
7E-9 Three balls are randomly dropped into three boxes, where any ball is
equally likely to fall into each box. Specify an appropriate sample space and
determine the probability that exactly one box will be empty.
7E-10 An electronic system has four components labeled as 1, 2, 3, and 4.
The system has to be used during a given time period. The probability
that component i will fail during that time period is f
i
for i = 1, . . . , 4.
Failures of the components are physically independent of each other. A
system failure occurs if component 1 fails or if at least two of the other
components fail. Specify an appropriate sample space and determine the
probability of a system failure.
7E-11 The Manhattan distance of a point (x, y) in the plane to the origin
(0, 0) is dened as |x| +|y|. You choose at random a point in the unit square
{(x, y) : 0 x, y 1}. What is the probability that the Manhattan distance
of this point to the point (0, 0) is no more than a for 0 a 2?
7E-12 You choose at random a point inside a rectangle whose sides have the
lengths 2 and 3. What is the probability that the distance of the point to the
closest side of the rectangle is no more than a given value a with 0 < a < 1?
2
7E-13 Pete tosses n + 1 fair coins and John tosses n fair coins. What is the
probability that Pete gets more heads than John? Answer this question rst
for the cases n = 1 and n = 2 before solving the general case.
7E-14 Three desperados A, B and C play Russian roulette in which they take
turns pulling the trigger of a six-cylinder revolver loaded with one bullet.
Each time the magazine is spun to randomly select a new cylinder to re as
long the deadly shot has not fallen. The desperados shoot according to the
order A, B, C, A, B, C, . . .. Determine for each of the three desperados the
probability that this desperado will be the one to shoot himself dead.
7E-15 A fair coin is tossed 20 times. The probability of getting the three
or more heads in a row is 0.7870 and the probability of getting three or
more heads in a row or three or more tails in a row is 0.9791.
2
What is the
probability of getting three or more heads in a row and three or more tails
in a row?
7E-16 The probability that a visit to a particular car dealer results in neither
buying a second-hand car nor a Japanese car is 55%. Of those coming to the
dealer, 25% buy a second-hand car and 30% buy a Japanese car. What is
the probability that a visit leads to buying a second-hand Japanese car?
7E-17 A fair die is repeatedly rolled and accumulating counts of 1s, 2s, . . .,
6s are recorded. What is an upper bound for the probability that the six
accumulating counts will ever be equal?
7E-18 A fair die is rolled six times. What is the probability that the largest
number rolled is r for r = 1, . . . , 6?
7E-19 Mr. Fermat and Mr. Pascal are playing a game of chance in a cafe
in Paris. The rst to win a total of ten games is the overall winner. Each
of the two players has the same probability of
1
2
to win any given game.
Suddenly the competition is interrupted and must be ended. This happens
at a moment that Fermat has won a games and Pascal has won b games with
a < 10 and b < 10. What is the probability that Fermat would have been
the overall winner when the competition would not have been interrupted?
Hint: imagine that another 10a+10b1 games would have been played.
7E-20 A random number is repeatedly drawn from 1, 2, . . . , 10. What is the
probability that not all of the numbers 1, 2, . . . , 10 show up in 50 drawings?
7E-21 Three couples attend a dinner. Each of the six people chooses ran-
domly a seat at a round table. What is the probability that no couple sits
together?
2
These probabilities can be computed by using an absorbing Markov chain discussed
in Chapter 15 of the book.
3
7E-22 You roll a fair die six times. What is the probability that three of the
six possible outcomes do not show up and each of the other three possible
outcomes shows up two times? What is the probability that some outcome
shows up at least three times?
7E-23 In a group of n boys and n girls, each boy chooses at random a girl
and each girl chooses at random a boy. The choices of the boys and girls are
independent of each other. If a boy and a girl have chosen each other, they
form a couple. What is the probability that no couple will be formed?
7E-24 Consider a communication network with four nodes n
1
, n
2
, n
3
and n
4
and ve directed links l
1
= (n
1
, n
2
), l
2
= (n
1
, n
3
), l
3
= (n
2
, n
3
), l
4
= (n
3
, n
2
),
l
5
= (n
2
, n
4
) and l
6
= (n
3
, n
4
). A message has to be sent from the source
node n
1
to the destination node n
4
. The network is unreliable. The proba-
bility that the link l
i
is functioning is p
i
for i = 1, . . . , 5. The links behave
physically independent of each other. A path from node n
1
to node n
4
is only
functioning if each of its links is functioning. Use the inclusion-exclusion for-
mula to nd the probability that there is some functioning path from node
n
1
to node n
4
. How does the expression for this probability simplify when
p
i
= p for all i?
Chapter 8
8E-1 Three fair dice are rolled. What is the probability that the sum of the
three outcomes is 10 given that the three dice show dierent outcomes?
8E-2 A fair coin is tossed n times. What is the probability of heads on the
rst toss given that r heads were obtained in the n tosses?
8E-3 A hand of 13 cards is dealt from a standard deck of 52 cards. What is the
probability that it contains more aces than tens? How does this probability
change when you have the information that the hand contains at least one
ace?
8E-4 In a high school class, 35% of the students take Spanish as a foreign
language, 15% take French as a foreign language, and 40% take at least one
of these languages. What is the probability that a randomly chosen student
takes French given that the student takes Spanish?
8E-5 Let A and B be independent events. Denote by the events A
c
and B
c
the complements of the events A and B. Verify that the events A and B
c
are independent. Conclude directly from this result that the events A
c
and
B
c
are also independent.
8E-6 Fifty dierent numbers are arranged in a matrix with 5 rows and 8
columns. You pick at random one number from the matrix. Let A be the
4
event that the number comes from an odd-numbered row and B be the event
that the number comes from the rst ve columns. Are the events A and B
independent?
8E-7 Consider again the Problems 7E-3, 7E-7 and 7E-8. Use conditional
probabilities to solve these problems.
8E-8 A bowl contains four red and four blue balls. As part of drawing lots,
you choose four times two balls at random from the bowl without replace-
ment. What is the probability that one one red and one blue ball are chosen
each time?
8E-9 You are among N players that will play a competition. A lottery is
used to determine the placement of each player. You have an advantage.
Two tickets with your name are put in a hat, while for each of the other
players only one ticket with her/his name is put in the hat. The hat is well
shaken and tickets are drawn one by one from the hat. The order of names
appearing determines the placement of each player. What is the probability
that you will get assigned the nth placement for n = 1, 2, . . . , N?
8E-10 Twenty-ve people choose each at random a number from 1, 2, . . . , 100,
independently of each other. Next the chosen numbers are announced one
by one. The rst person (if any) who announces a number that has been
announced before wins a bonus. Which person has the largest probability to
win the bonus?
8E-11 In a poker game with three players A, B and C, the dealer is chosen
by the following procedure. In the order A, B, C, A, B, C, . . ., a card from
a well-shued deck is dealt to each player until someone gets an ace. This
rst player receiving an ace gets to start the game as dealer. Do you think
that everyone has an equal chance to become dealer?
8E-12 A drunkard removes two randomly chosen letters of the message HAPPY
HOUR that is attached on a billboard in a pub. His drunk friend puts the
two letters back in a random order. What is the probability that HAPPY
HOUR appears again?
8E-13 A professor gives only two types of exams, easyand hard. You
will get a hard exam with probability 0.80. The probability that the rst
question on the exam will be marked as dicult is 0.90 if the exam is hard
and is 0.15 otherwise. What is the probability that the rst question on your
exam is marked as dicult. What is the probability that your exam is hard
given that the rst question on the exam is marked as dicult?
8E-14 Player 1 tosses N + 1 times a fair coin and player 2 tosses N times a
fair coin. Player 1 wins the game if player 1 tosses more heads than player
5
2; otherwise, player 2 wins.
(a) What is the probability of a tie after N tosses?
(b) What is the probability that player 1 wins the game?
8E-15 A jar contains ve blue balls and ve red balls. You roll a fair die
once. Next you randomly draw (without replacement) as many balls from
the jar as the number of points you have rolled with the die.
(a) What is the probability that all of the balls drawn are blue?
(b) What is the probability that the number of points shown by the die is r
given that all of the balls drawn are blue?
8E-16 A tennis tournament is arranged for 8 players. It is organized as a
knockout tournament. First, the 8 players are randomly allocated over four
groups of two players each. In the semi-nals the winners of the groups 1
and 2 meet each other and the winners of the groups 3 and 4. In any match
either player has a probability 0.5 of winning. John and Pete are among the 8
players. What is the probability that they meet each other in the semi-nals?
What is the probability that they meet each other in the nal?
8E-17 Consider again Problem 7E-19. Show how this problem can be solved
by a recursive approach.
8E-18 A biased coin is tossed repeatedly. The probability that a toss of the
coin results in heads is p with 0 < p < 1.
(a) Give a recursion for the probability that the total number of heads after
n tosses is even.
(b) Give a recursion for the probability that a sequence of n tosses does not
show ve or more consecutive heads.
8E-19 A lottery organization distributes one million tickets every week. At
one end of the ticket, there is a visible printed number consisting of six
digits, say 070469. At the other end of the ticket, another six-digit number
is printed, but this number is hidden by a layer of scratch-away silver paint.
The ticket holder scratches the paint away to reveal the underlying number.
If the number is the same as the number at the other end of the ticket, it
is a winning ticket. The two six-digit numbers on each of the one million
tickets printed each week are randomly generated in such a way that no
two tickets are printed with the same visible numbers or the same hidden
numbers. Assume that in a particular week only one half of the tickets
printed are sold. What is the probability of exactly r winners in that week
for r = 0, 1, . . .?
8E-20 In a binary transmission channel, a 1 is transmitted with probability
0.8 and a 0 with probability 0.2. The conditional probability of receiving a
6
1 given that a 1 was sent is 0.95, the conditional probability of receiving a 0
when a 0 was sent is 0.99. What is the probability that a 1 was sent when
receiving a 1? Uses Bayes formula in odds form to answer this question.
8E-21 On the island of liars each inhabitant lies with probability
2
3
. You
overhear an inhabitant making a statement. Next you ask another inhabitant
whether the inhabitant you overheard spoke truthfully. Use Bayes rule in
odds form to nd the probability that the inhabitant you overheard indeed
spoke truthfully given that the other inhabitant says so.
8E-22 An oil explorer performs a seismic test to determine whether oil is
likely to be found in a certain area. The probability that the test indicates
the presence of oil is 90% if oil is indeed present in the test area and the
probability of a false positive is 15% if no oil is present in the test area.
Before the test is done, the explorer believes that the probability of presence
of oil in the test area is 40%. Use Bayes rule in odds form to revise the
value of the probability of oil being present in the test area given that the
test gives a positive signal.
8E-23 An isolated island is ruled by a dictator. Every family on the island
has two children. Each child is equally likely a boy or a girl. The dictator has
decreed that each rst girl (if any) born to the family must bear the name
Mary Ann (the name of the beloved mother-in-law of the dictator). Two
siblings never have the same name. You are told that a randomly chosen
family that is unknown to you has a girl named Mary Ann. What is the
probability that this family has two girls?
The dictator has passed away. His son, a womanizer, has changed the
rules. For each rst girl born to the family a name must be chosen at random
from 10 specic names including the name Mary Ann, while for each second
girl born to the family a name must be randomly chosen from the remaining
9 names. What is now the probability that a randomly chosen family has two
girls when you are told that this family has a girl named Mary Ann? Can
you intuitively explain why this probability is not the same as the previous
probability?
8E-24 A family is chosen at random from all three-child families. What is
the probability that the chosen family has one boy and two girls if the family
has a boy among the three children? Use Bayes rule in odds form to answer
this question.
8-25 (a) A box contains 10,000 coins. One of the coins has heads on both
sides but all the other coins are fair coins. You choose at random one of the
coins. Use Bayes rule in odds form to nd the probability that you have
chosen the two-headed coin given that the rst 15 tosses all have resulted in
7
heads. What is the answer when you would have obtained 25 heads in a row
in the rst 25 tosses?
(b) A box contains r + 1 coins i = 0, 1, . . . , r. Coin i lands heads with
probability
i
r
for i = 0, 1, . . . , r. You choose at random one of the coins. Use
Bayes rule in odds form to nd the probability that you have chosen coin s
given that each of the rst n tosses has resulted in heads.
8E-26 Your friend has generated two random numbers from 1, . . . , 10, inde-
pendently of each other. Use Bayesrule in odds form to answer the following
two questions.
(a) What is the probability that both numbers are even given the information
that there is an even number among the two numbers?
(b) What is the probability that both numbers are even given the information
that the number 2 is among the two numbers?
8E-27 Your friend has fabricated a loaded die. In doing so, he has rst
simulated a number at random from 0.1, 0.2, 0.3, and 0.4. He tells you that
the die is loaded in such a way that any roll of the die results in the outcome
6 with a probability which is equal to the simulated number. Next the die is
rolled 300 times and you are informed that the outcome 6 has appeared 75
times. What is the posterior distribution of the probability that a single roll
of the die gives a 6?
8E-28 Your friend is a basketball player. To nd out how good he is in free
throws, you ask him to shoot 10 throws. You assume the three possible values
0.25, 0.50 and 0.75 for the success probability of the free shots of your friend.
Before the 10 throws are shot, you believe that these three values have the
respective probabilities 0.2, 0.6 and 0.2. What is the posterior distribution
of the success probability given that your friend scores 7 times out of the 10
throws?
Chapter 9
9E-1 Five men and ve women are ranked according to their scores on an
exam. Assume that no two scores are the same and all possible rankings are
equally likely. Let the random variable X be the highest ranking achieved
by a women. What is the probability mass function of X?
9E-2 You roll a fair dice twice. Let the random variable X be the product of
the outcomes of the two rolls. What is the probability mass function of X?
What are the expected value and the standard deviation of X?
9E-3 In a lottery a four-digit number is chosen at random from the range
0000 9999. A lottery ticket costs $2. You win $50 if your ticket matches
8
the last two digits but not the last three, $500 if your ticket matches the last
three digits but not all four, and $5,000 if your ticket matches all four digits.
What is the expected payo on a lottery ticket? What is the house edge of
the lottery?
9E-4 The following game is oered. There are 10 cards face-down numbered
1 through 10. You can pick one card. Your payo is $0.50 if the number on
the card is less than 5 and is the dollar value on the card otherwise. What
are the expected value and the standard deviation of your payo?
9E-5 A fair die is rolled six times. What are the expected value and the
standard deviation of the smallest number rolled?
9E-6 Eleven closed boxes are put in random order in front of you. The boxes
1, . . . , 10 contain a
1
, . . . , a
10
dollars, but box 11 contains a devils penny. You
may open as many boxes as you wish, but they must be opened one by one.
You can keep the money from the boxes you have opened as long as you
have not opened the box with the devils penny. Once you open this box, the
game is over and you lose all the money gathered so far. What is a reasonable
stopping rule to maximize the expected value of your gain?
9E-7 You play a sequence of s games, where s 2 is xed. The outcomes of
the various games are independent of each other. The probability that you
will win the kth game is
1
k
for k = 1, 2, . . . , s. You get one dollar each time
you win two games in a row. What is the expected value of the total amount
you will get?
9E-8 You toss a biased coin with probability p of heads, while your friend
tosses at the same time a fair coin. What is the probability distribution of the
number of tosses until both coins simultaneously show the same outcome?
9E-9 You distribute randomly 25 apples over 10 boxes. What is the expected
value of the number of boxes that will contain exactly k apples for k =
0, 1, . . . , 25?
9E-10 You have a thoroughly shued deck of 52 cards. Each time you choose
one card from the deck. The drawn card is put back in the deck and all
52 cards are again thoroughly shued. You continue this procedure until
you have seen all four dierent aces. What are the expected value and the
standard deviation of the number of times you have to draw a card until you
have seen all four dierent aces?
9E-11 A group of m people simultaneously enter an elevator at the ground
oor. Each person randomly chooses one of the r oors 1, 2, . . . , r as the exit
oor, where the choices of the persons are independent of each other. The
elevator only stops on a oor if at least one person wants to exit on that oor.
9
No other people enter the elevator at any of the oors 1, 2, . . . , r. What are
the expected value and the standard deviation of the number of stops the
elevator will make?
9E-12 (a) A random integer is repeatedly drawn from 1, 2, . . . , 10. What are
the expected value and the standard deviation of the number of integers from
1, 2, . . . , 10 that do not show up in 20 drawings?
(b) In each drawing of the Lotto 6/45 six dierent integers are randomly
chosen from 1, 2, . . . , 45. What are the expected value and the standard
deviation of the number of integers from 1, 2, . . . , 45 that do not show up in
15 drawings?
9E-13 Take a random permutation of the integers 1, 2, . . . , n. Let us say that
the integers i and j with i = j are switched if the integer i occupies the jth
position in the random permutation and the integer j the ith position. What
is the expected value of the total number of switches?
9E-14 Suppose you know that the hands of you and your bridge partner
contain eight of the 13 spades in the deck. What is the probability of 3-2
split of the remaining ve spades in the bridge hands of your opponents?
9E-15 You choose at random an integer from 1, 2, . . . , 6. Next you roll a fair
die until you get an outcome that is larger than or equal to the randomly
chosen integer. What is the probability mass function of the number of times
you will roll the die? What are the expected value and the standard deviation
of the number of times you will roll the die?
9E-16 You have two coins. One coin is fair and the other is biased with
probability p of heads. The rst toss is done with the fair coin. At the
subsequent tosses the fair coin is used if the previous toss resulted in heads
and the biased coin is used otherwise. What is the expected value of the
number of heads in r tosses for r = 1, 2, . . .?
9E-17 Let X and Y be independent random variables, where X is binomially
distributed with parameters n and p and Y is binomially distributed with
parameters m and p.
(a) Explain in terms of Bernoulli experiments that X + Y is binomially
distributed with parameters n +m and p. Next give a formal proof.
(b) Verify that for xed k the probabilities P(X = j | X + Y = k) for
j = 0, . . . , k constitute a hypergeometric distribution.
9E-18 Let the number of events occurring in a given time interval be a Poisson
distributed random variable N with expected value . Each events is labeled
as a type-1 event with probability p and as a type-2 event with probability
1 p, independently of the other events. Let X be the number of type-1
10
events occurring in the time interval and Y be the number of type-2 events.
Prove that X and Y are Poisson distributed with expected values p and
(1 p) and that X and Y are independent. Hint: P(X = j, Y = k) =
P(AB), where the events A and B are given by A = {X = j, Y = k} and
B = {N = j +k}.
9E-19 (a) The random variable X is Poisson distributed with expected value
. Verify that E[g(X +1) Xg(X)] = 0 for any bounded function g(x) on
the integers 0, 1, . . . .
(b) Let X be a random variable on the integers 0, 1, . . . and > 0 a given
number. Prove that X has a Poisson distribution with expected value if
E[g(X + 1) Xg(X)] = 0 for any bounded function g(x) on the integers
0, 1, . . . . Hint: take the function g(x) dened by g(r) = 1 and g(x) = 0 for
x = r with r a xed nonnegative integer.
9E-20 You rst roll a fair die once. Next you roll the die as many times
as the outcome of this rst roll. Let the random variable X be the total
number of sixes in all the rolls of the die, including the rst roll. What is
the probability mass function of X?
9E-21 The following game is played in a particular carnival tent. There is
a box with 10 balls and four of those balls are gold-colored. You pay one
dollar to play the game. Then you draw blindly three balls from the box
(without replacement). Your get your original one-dollar stake back if you
draw exactly two gold-colored balls, while you win 10 dollars and you get your
original one-dollar stake back if you draw three gold-colored balls; otherwise,
you get nothing back. Is this a fair game?
9E-22 Accidentally, two depleted batteries got into a set of ve batteries.
To remove the two depleted batteries, the batteries are tested one by one
in a random order. Let the random variable X denote the number of bat-
teries that must be tested to nd the two depleted batteries. What is the
probability mass function of X?
9E-23 In a close election between two candidates A and B in a small town
the winning margin of candidate A is 1,422 to 1,405 votes. However, 101
votes are illegal and have to be thrown out. Assuming that the illegal votes
are not biased in any particular way and the count is otherwise reliable, what
is the probability the removal of the illegal votes changes the result of the
election?
9E-24 A bowl contains n red balls and m white balls. You randomly pick
without replacement one ball at a time until you have r red balls. What is
the probability that you need k draws?
11
9E-25 You and your friend both draw a random number from 1, 2, . . . , 10 at
the same time and independently of each other. This procedure is repeated
until you have drawn one of the four numbers 1, . . . , 4 or your friend has
drawn one of the six numbers 5, . . . , 10. The rst player to get one of his
marked numbers is the winner with the convention that you are the winner
if a tie occurs. What is your probability of winning the game? What is the
probability mass function of the length of the game?
3
9E-26 The G50 airplane is at the end of its lifetime. The remaining opera-
tional lifetime of the plane and is 3, 4 or 5 years each with probability
1
3
. A
decision must be made how many spare parts of a certain component to pro-
duce. The demand for spare parts of the component is Poisson distributed
with an expected value of units per year for each year of the remaining
lifetime of the plane, where the demands in the various years are independent
of each other. It is decided to produce Q units of the spare part. What is the
probability that the production size will not be enough to cover the demand?
What is the expected value of the shortage? What is the expected value of
the number of units left over at the end of the operational lifetime of the
plane?
9E-27 You have bought 10 young beech trees. They come from a garden
center and were randomly chosen from a collection of 100 trees consisting of
50 trees from tree-nurseryman A and 50 trees from tree-nurseryman B. Ten
percent of the trees from tree-nurseryman A and ve percent of the trees
from tree-nurseryman B do not know grow well. What is the probability
that no more than one of your ten trees will not grow well?
Chapter 10
10E-1 The density function of the continuous random variable X is given
by f(x) = c(x +

x) for 0 < x < 1 and f(x) = 0 otherwise. What is the


constant c? What is probability density of
1
X
?
10E-2 The radius of a circle is uniformly distributed on (0, 1). What is the
probability density of the area of the circle?
10E-3 You choose at random a point inside a rectangle whose sides have the
lengths 2 and 3. Let the random variable X be the distance from the point
to the closest side of the rectangle. What is the probability density of X?
What are the expected value and the standard deviation of X?
3
This problem is based on D. Neal and F. Polivka, A geometric gambling game, The
Mathematical Spectrum, Vol. 43 (2010), p. 20-25.
12
10E-4 Liquid waste produced by a factory is removed once a week. The
weekly volume of waste in thousands of gallons is a continuous random vari-
able with probability density function fx) = 105x
4
(1x)
2
for 0 < x < 1 and
f(x) = 0 otherwise. How to choose the capacity of a storage tank so that
the probability of overow during a given week is no more than 5%?
10E-5 Consider again Problem 10E-4. Assume that the storage tank has
a capacity of 0.9 expressed in thousands of gallons. The cost of removing
x > 0 units of waste at the end of the week is 1.25 + 0.5x. Additional costs
5 + 10z are incurred when the capacity of the storage tank is not sucient
and an overow of z > 0 units of waste occurs during the week. What are
the expected value and the standard deviation of the weekly costs?
10E-6 You have to make an one-time business decision how much stock to
order in order to meet a random demand during a single period. The demand
is a continuous random variable X with a given probability density f(x).
Suppose you decide to order Q units. What is the probability that the initial
stock Q will not be enough to meet the demand? What is the expected value
of the stock left over at the end of the period? What is the expected value
of demand that cannot be satised from stock?
10E-7 Let Q be a xed point on the circumference of a circle with radius
r. Choose at random a point P on the circumference of the circle and let
the random variable X be the length of the line segment between P and Q.
What are the expected value and the standard deviation of X?
10E-8 An insurance policy for water damage pays an amount of damage up
to $450. The amount of damage is uniformly distributed between $250 and
$1,250. The amount of damage exceeding $450 is covered by a supplement
policy up to $500. Let the random variable Y be the amount of damage paid
by the supplement policy. What are the expected value and the probability
distribution function of Y ?
10E-9 An expensive item is being insured against early failure. The lifetime
of the item is normally distributed with an expected value of seven years
and a standard deviation of two years. The insurance will pay a dollars if
the item fails during the rst or second year and
1
2
a dollars if the item fails
during the third or fourth year. If a failure occurs after the fourth year, then
the insurance pays nothing. How to choose a such that the expected value
of the payment per insurance is $50?
10E-10 Let be a randomly chosen angle in (0,

4
). The random variable Y
is dened as the y-coordinate of the point at which the ray through the origin
at angle intersects the line x = 1 in the plane. What are the expected
13
value and the standard deviation of the area of the triangle with the corner
points (0, 0), (1, 0) and (1, Y )? What is the probability density of this area
of this triangle?
10E-11 A shot is red at a very large circular target. The horizontal and
vertical coordinates of the point of impact are independent random variables
each having a standard normal density. Here the center of the target is taken
as the origin. What is the density function of the distance from the center
of the target to the point of impact? What are the expected value and the
mode of this distance?
10E-12 Let X and Y be independent random variables each having the stan-
dard normal distribution. Consider the circle centered at the origin and
passing through the point (X, Y ). What is the probability density of the
area of the circle? What is the expected value of this area?
10E-13 A service station has a slow server (server 1) and a fast server (server
2). Upon arrival at the station, you are routed to server i with probability p
i
for i = 1, 2, where p
1
+ p
2
= 1. The service time at server i is exponentially
distributed with parameter
i
for i = 1, 2. What is the probability density
of your service time at the station?
10E-14 Use rst principles to answer the following questions.
(a) The random variable X has a standard normal distribution. What is the
probability density of the random variable Y = X
2
?
(b) The random variable X has a standard normal distribution. What is the
probability density of the random variable Y =
_
|X|?
10E-15 The random variable X has the probability density function f(x) =
8

_
x(1 x) for 0 < x < 1 and f(x) = 0 otherwise. What is the probability
density function of the random variable Y = 2X + 1?
10E-16 The random variable X has the Cauchy density f(x) =
1
(1+x
2
)
for
< x < . Use the transformation rule for random variables to nd the
probability density of the random variable Y =
1
X
.
10E-17 Let X be a continuous random variable with probability density func-
tion f(x). How would you dene the conditional expected value of X given
that X a? What is E(X | X a) when X is exponentially distributed
with parameter ?
10E-18 You wish to cross a one-way trac road on which cars drive at a
constant speed and pass according to independent interarrival times having
an exponential distribution with an expected value of 1/ seconds. You can
only cross the road when no car has come round the corner since c time
seconds. What is the probability distribution of the number of passing cars
14
before you can cross the road when you arrive at an arbitrary moment? What
property of the exponential distribution do you use?
10E-19 The amount of time needed to wash a car at a car washing station is
exponentially distributed with an expected value of 15 minutes. You arrive
at a car washing station, while the washing station is occupied and one other
car is waiting for a washing. The owner of this car informs you that the car in
the washing station is already there for 10 minutes. What is the probability
that the car in the washing station will need no more than ve other minutes?
What is the probability that you have to wait more than 20 minutes before
your car can be washed?
10E-20 You simulate a random observation from an exponentially distributed
random variable X with expected value 1. What is the probability that the
closest integer to the random observation is odd? What is this probability
if the random observation is larger than a given even integer r? Can you
explain why the two probabilities are the same?
10E-21 In a video game with a time slot of xed length T, signals are gen-
erated according to a Poisson process with rate , where T >
1

. During the
time slot you can push a button only once. You win if at least one signal
occurs in the time slot and you push the button at the occurrence of the last
signal. Your strategy is to let pass a xed time s with 0 < s < T and push
the button upon the rst occurrence of a signal (if any) after time s. What
is your probability of winning the game? What value of s maximizes this
probability?
10E-22 Cars pass through an out-of-the way village according to a Poisson
process. The probability of one or more cars passing through the village
during one hour is 0.64. What is the probability of a car passing through the
village during the next half hour?
10E-23 Two instruments are used to measure the unknown length of a beam.
If the true length of the beam is l, the measurement error made by the
rst instrument is normally distributed with mean 0 and standard devia-
tion 0.006l and the the measurement error made by the rst instrument is
normally distributed with mean 0 and standard deviation 0.004l. The two
measurement errors are independent of each other. What is the probability
that the average value of the two measurements is within 0.5% of the actual
length of the beam?
10E-24 The lifetimes of two components in an electronic system are inde-
pendent random variables X
1
and X
2
, where X
i
has a normal distribution
with an expected value of
i
time units and a standard deviation of
i
time
15
units. What is the probability that the lifetimes of the two components
expire within a time units from each other?
10E-25 A space mission will take 150 days. A number of copies of a daily-use
appliance must be taken along. The amount of time the appliance can be
used is exponentially distributed with an expected value of two days. What
is the probability mass function of the number of copies of the appliance
to be used when an innite supply would be available? Use the normal
approximation to nd how many copies of the appliance should be stocked
so that the probability of a shortage during the mission is no more than 10
3
.
Compare the approximate result to the exact result.
10E-26 A new casino has just been opened. The casino owner makes the
following promotional oer to induce gamblers to play at his casino. People
who bet $10 on red get half their money back if they lose the bet, while they
get the usual payout of $20 if they win the bet. This oer applies only to the
rst 2,500 bets. In the casino European roulette is played so that a bet on
red is lost with probability
19
37
and is won with probability
18
37
. Use the normal
distribution to approximate the probability that the casino owner will lose
more than 6,500 dollars on the promotional oer.
10E-27 Consider again Problem 10E-26. Assume now that the casino owner
makes an oer only to the most famous gambler in town. The casino owner
lends $1,000 to the gambler and proposes him to make 100 bets on red with
this starting capital. The gambler is allowed to stake any amount of his
bankroll at any bet. The gambler gets one fourth of the staked money back
if he loses the bet, while he gets double the staked money back if he wins
the bet. As reward, the gambler can keep to himself any amount in excess
of $1,000 when stopping after 100 bets. Suppose that the gambler decides
to stake each time a xed percentage of 5% of his current bankroll. Use the
normal distribution to approximate the probability that the gambler takes
home more than d dollars for d = 0, 500, 1,000 and 2,500. Hint: consider
the logarithm of the size of the bankroll of the gambler.
4
10E-28 The failure rate function of the lifetime of a vacuum tube is r(x) =
(
1
e

1
x
+
2
e

2
x
(
1
+
2
)e
(
1
+
2
)x
)
_
(e

1
x
+e

2
x
e
(
1
+
2
)x
) for x > 0,
where 0 <
1
<
2
. Verify that the function r(x) has a bathtub shape and
determine the corresponding probability distribution function.
Chapter 11
4
This problem is related to Kelly betting discussed in the Sections 2.7 and 5.9 of the
book.
16
11E-1 (a) A fair coin is tossed three times. Let X be the number of heads
among the rst two tosses and Y be the number of heads among the last two
tosses. What is the joint probability mass function of X and Y ? What is
E(XY )?
(b) You have two fair coins. The rst coin is tossed ve times. Let the
random variable X be the number of heads showing up in these ve tosses.
The second coin is tossed X times. Let Y be the number of heads showing up
in the tosses of the second coin. What is the joint probability mass function
of X and Y ? What is E(X +Y )?
11E-2 In the nal of the World Series Baseball, two teams play a series
consisting of at most seven games until one of the two teams has won four
games. Two unevenly matched teams are pitted against each other and the
probability that the weaker team will win any given game is equal to 0.45.
Let X be equal to 1 if the stronger team is the overall winner and X be equal
to 0 otherwise. The random variable Y is dened as the number of games
the nal will take. What is the joint probability mass function of X and Y ?
11E-3 You roll a fair die once. Let N be the outcome of this roll. Each of
two persons toss N fair coins, independently of each other. Let X be the
number of heads obtained by the rst person and Y be the number of heads
obtained by the second person. What is the joint probability mass function
of X and Y ? What is the numerical value of P(X = Y )?
11E-4 You simultaneously roll d fair dice. Let the random variable X be
the outcome of the highest scoring die and Y be the outcome of the second-
highest scoring die with the convention that the second-highest score equals
the highest score in the case that two or more dice yield the highest score.
What is the joint probability mass function of X and Y ?
11E-5 The joint probability mass function of the lifetimes X and Y of two
connected components in a machine can be modeled by p(x, y) =
e
2
x!(yx)!
for
x = 0, 1, . . . and y = x, x + 1, . . ..
(a) What are the marginal distributions of X and Y ?
(b) What is the joint probability mass function of X and Y X? Are X and
Y X independent?
(c) What is the correlation between X and Y ?
11E-6 A fair coin is rolled six times. Let X be the number of times a 1
is rolled and Y be the number of times a 6 is rolled. What is the joint
probability mass function of X and Y ? What is the correlation coecient of
X and Y ?
11E-7 The joint density function of the continuous random variables X and
17
Y is given by f(x, y) = cxy for 0 < y < x < 1 and f(x, y) = 0 otherwise.
What is the constant c? What are the marginal densities f
X
(x) and f
Y
(y)?
11E-8 The joint density function of the random variables X and Y is given
by f(x, y) = x +y for 0 x, y 1 and f(x, y) = 0 otherwise. Consider the
circle centered at the origin and passing through the point (X, Y ). What is
the probability that the circumference of the circle is no more than 2?
11E-9 There are two alternative routes for a ship passage. The sailing times
for the two routes are random variables X and Y that have the joint density
function f(x, y) =
1
10
e

1
2
(y+3x)
for 5 < x < 10, y > x 3 and f(x, y) = 0
otherwise. What is P(X < Y )?
11E-10 The joint density function of the random variables X and Y is given
by f(x, y) = xe
x(y+1)
for x, y > 0 and f(x, y) = 0 otherwise. What is the
density function of the random variable XY ?
11E-11 The joint density function of the random variables X and Y is given
by f(x, y) =
1
2
(x +y)e
(x+y)
for x, y > 0 and f(x, y) = 0 otherwise. What is
the density function of the random variable X +Y ?
11E-12 The lifetimes X and Y of two components in a machine have the
joint density function f(x, y) =
1
4
(2y + 2 x) for 0 < x < 2, 0 < y < 1 and
f(x, y) = 0 otherwise.
(a) What is the probability density of the time until neither of two compo-
nents is still working?
(b) What is the probability distribution of the amount of time that the life-
time X survives the lifetime Y ?
11E-13 (a) The joint density function f(a, b) of the random variables A and
B is given by f(a, b) = a+b for 0 < a, b < 1 and f(a, b) = 0 otherwise. What
is the probability that the equation Ax
2
+Bx + 1 = 0 has two real roots?
(b) The joint density function f(a, b, c) of the random variables A, B, and
C is given by f(a, b, c) =
2
3
(a + b + c) for 0 < a, b, c < 1 and f(a, b, c) = 0
otherwise. What is the probability that the equation Ax
2
+Bx +C = 0 has
two real roots?
11E-14 An unreliable electronic system has two components hooked up in
parallel. The lifetimes X and Y of the two components have the joint density
function f(x, y) = e
(x+y)
for x, y 0. The system goes down when both
components have failed. The system is inspected every T time units. What
is the probability that the system goes down between two inspections? What
is the expected amount of time the system is down between two inspections?
11E-15 An electronic device contains two circuits. The second circuit is a
backup for the rst and is switched on only when the rst circuit has failed.
18
The electronic device goes down when the second circuit fails. The continuous
random variables X and Y denote the lifetimes of the rst circuit and the
second circuit and have the joint density function f(x, y) = 24/(x + y)
4
for x, y > 1 and f(x, y) = 0 otherwise. What is the expected value of the
time until the electronic device goes down? What is the probability density
function of this time?
11E-16 Choose three random numbers X
1
, X
2
and X
3
from (0, 1), indepen-
dently of each other. What is the probability P(X
1
> X
2
+ X
3
)? What is
the probability that the largest of the three random numbers is greater than
the sum of the other two?
11E-17 Let X
1
, X
2
, . . . , X
n
be independent random variables that are uni-
formly distributed on (0, 1). What is P(X
1
+ X
2
+ + X
n
1)? Answer
this question rst for n = 2 and n = 3.
11E-18 The random variables X andY are independent and uniformly dis-
tributed on (0, 1). Let V = X +Y and W =
X
Y
. What is the joint density of
V and W? Are V and W independent?
11E-19 The random variables V and W are dened by V = Z
2
1
+ Z
2
2
and
W = Z
2
1
Z
2
2
, where Z
1
and Z
2
are independent random variables each
having the standard normal distribution. What is the joint density function
of V and W? Are V and W independent?
11E-20 The random variables X andY are independent and exponentially
distributed with parameter . Let V = X + Y and W =
X
X+Y
. What is the
joint density of V and W? Prove that V and W are independent.
11E-21 The continuous random variables X and Y have the joint density
function f(x, y) = cxe

1
2
x(1+y
2
)
for x, y > 0 and f(x, y) = 0 otherwise.
What is the constant c? What are the marginal densities of X and Y ? Show
that the random variables Y

X and X are independent. Hint: Use the fact


that the gamma density

x
1
e
x
/() integrates to 1 over (0, ) for any
, > 0 and note that (1.5) =
1
2

.
11E-22 The continuous random variables X and Y have the joint density
function f(x, y) = 6(x y) for 0 < y < x < 1 and f(x, y) = 0 otherwise.
Determine the correlation coecient of X and Y .
11E-23 John and Pete are going to throw the discus in their very last turn at
a champion game. The distances thrown by John and Pete are independent
random variables D
1
and D
2
that are N(
1
,
2
1
) and N(
2
,
2
2
) distributed.
What the best linear prediction of the distance thrown by John given that
the dierence between the distances of the throws of John and Pete is d?
19
Month Observed Expected
Jan 60,179 61,419.5
Feb 54,551 55,475.7
Mar 59,965 61,419.5
Apr 57,196 59,438.2
May 59,444 61,419.5
Jun 59,459 59,438.2
Jul 62,166 61,419.5
Aug 60,598 61,419.5
Sep 62,986 59,438.2
Oct 64,542 61,419.5
Nov 60,745 59,438.2
Dec 61,334 61,419.5
Chapter 12
12E-1 Let Z
1
and Z
2
be independent random variables each having the
standard normal distribution. Dene the random variables X and Y by
X = Z
1
+3Z
2
and Y = Z
1
+Z
2
. Argue that the joint distribution of (X, Y )
is a bivariate normal distribution. What are the parameters of this distribu-
tion?
12E-2 Let the random vector (X, Y ) have the standard bivariate normal
distribution with correlation coecient = 0.5. What are the values of a for
which the random variables V = aX+Y and W = X+aY are independent?
12E-3 Let the random vector (X, Y ) have the standard bivariate normal dis-
tribution with correlation coecient with 1 < < 1. Let Z be an N(0, 1)
distributed random variable that is independent of the random variable X.
Verify that the random vector (X, X +
_
1
2
Z) has the same standard
bivariate normal distribution as (X, Y ).
12E-4 Let the random vector (X, Y ) have the standard bivariate normal
distribution with correlation coecient with 1 < < 1. What are the
probabilities P(Y > X | X > 0) and P(Y/X 1)?
12E-5 Let X = (X
1
, . . . , X
n
) and Y = (Y
1
, . . . , Y
m
) be independent ran-
dom vectors each having a multivariate normal distribution. Prove that the
random vector (X, Y) has also a multivariate normal distribution.
12E-6 In the table the column Observed gives the observed frequencies of
birth months for the children born in England and Wales in the year 2010.
The column Expected gives how many births could have been expected in
20
each month under the hypothesis that all birth dates are equally likely. Use
the chi-square test to make clear that the assumption of equally likely birth
dates is not satised in reality.
5
12E-7 In a famous physics experiment performed by Rutherford, Chadwick
and Ellis in 1920, the number -particles emitted by a piece of radioactive
material were counted during 2,608 time intervals of each 7.5 seconds. There
were 57 intervals with zero particles, 203 intervals with 1 particle, 383 in-
tervals with 2 particles, 525 intervals with 3 particles, 532 intervals with 4
particles, 408 intervals with 5 particles, 273 intervals with 6 particles, 139
intervals with 7 particles, 45 intervals with 8 particles, 27 intervals with 9
particles, 10 intervals with 10 particles, 4 intervals with 11 particles, 0 in-
tervals with 12 particles, 1 interval with 13 particles, and 1 interval with
14 particles. Use a chi-square test to investigate how closely the observed
frequencies conform to Poisson frequencies.
12E-8 Vacancies in the U.S. Supreme Court over the 78-years period 1933-
2010 have the following history: 48 years with 0 vacancies, 23 years with 1
vacancy, 7 years with 2 vacancies, and 0 years with 3 vacancies. Test the
hypothesis that the number of vacancies per year follows a Poisson distribu-
tion.
Chapter 13
13E-1 In the nal of the World Series Baseball, two teams play a series
consisting of at most seven games until one of the two teams has won four
games. Two unevenly matched teams are pitted against each other and the
probability that the weaker team will win any given game is equal to 0.45.
What is the conditional probability mass function of the number of games
played in the nal given that the weaker team has won the nal?
13E-2 Let P(X = n, Y = k) =
_
n
k
_ _
1
6
_
k
_
1
3
_
nk
for k = 0, 1, . . . , n and
n = 1, 2, . . . be the joint probability mass function of the random variables
X and Y . What is the conditional distribution of Y given that X = n?
13E-3 A fair die is repeatedly rolled. Let the random variable X be the
number of rolls until the face value 1 appears and Y be the number of rolls
until the face value 6 appears. What are E(X | Y = 2) and E(X | Y = 20)?
13E-4 The continuous random variables X and Y satisfy f
Y
(y | x) =
1
x
for
0 < y < x and f
Y
(y | x) = 0 otherwise. The marginal density function of X
5
For birthday problems it can be shown the probability of matching birthdays gets
larger when the assumption that each day of the year is equally likely as birthday is
dropped.
21
is given by f
X
(x) = 2x for 0 < x < 1 and f
X
(x) = 0 otherwise. What is the
conditional density f
X
(x | y)? What is E(X | Y = y)?
13E-5 The random variables X and Y have the joint density function f(x, y) =
e
y
for 0 x y and f(x, y) = 0 otherwise. Describe a method to simulate
a random observation from f(x, y).
13E-6 Consider again Problem 11E-15. What is the expected value of the
lifetime of the second circuit given that the rst circuit has failed after s time
units? What is the probability that the second circuit will work more than
v time units given that the rst circuit has failed after s time units?
13E-7 Eleven closed boxes are put in random order in front of you. The boxes
1, . . . , 10 contain a
1
, . . . , a
10
dollars, but box 11 contains a devils penny. You
may mark as many boxes as you wish. The marked boxes are opened. You
win the money from these boxes if the box with the devils penny is not
among the opened boxes; otherwise, you win nothing. How many boxes
should you mark to maximize your expected return?
13E-8 Let s be a xed number with 0 < s < 1. By choosing at random
a point U in the interval (0, 1), this interval is divided in two subintervals.
What is the expected length of the subinterval covering the number s?
13E-9 (a) You rst toss a fair coin ve times. Next you toss the coin as many
times as the number of heads showing up in these ve tosses. Let the random
variable X be the number of heads in all tosses of the coin, including the rst
ve tosses. Use conditional expectations to nd the expected value of X.
(b) You rst roll a fair die once. Next you roll the die as many times as the
outcome of this rst roll. Let the random variable X be the total number
of sixes in all the rolls of the die, including the rst roll. Use conditional
expectations to nd the expected value of X.
13E-10 Let X
1
and X
2
be independent random variables each having a ge-
ometric distribution with parameter p. What is the conditional probability
mass function of X
1
given that X
1
+X
2
= r?
13E-11 You draw at random a number p from the interval (0, 1). Next you
toss n times a coin with probability p of heads. What is the probability mass
function of the number of times that heads will appear? Hint: Use the fact
that the beta integral
_
1
0
x
r1
(1x)
s1
dx is equal to (r1)!(s1)!/(r+s1)!
for positive integers r and s.
13E-12 Let X
1
and X
2
be independent random variables each having an ex-
ponential distribution with parameter . What is the conditional probability
density function of X
1
given that X
1
+X
2
= s?
22
13E-13 Let and R be independent random variables, where is uniformly
distributed on (, ) and R is a positive random variable with density
function re

1
2
r
2
for r > 0. Dene the random variables V and W by V =
Rcos() and W = Rsin(). What is the conditional density function of V
given that W = w? What is E(V | W = w)?
13E-14 Let X
1
, X
2
, . . . be independent random variables that are uniformly
distributed on (0, 1). The random variable N is dened as the smallest n 2
for which X
n
> X
1
. What is the probability mass function of N?
13E-15 Let U
1
and U
2
be two independent random variables that are uni-
formly distributed on (0, 1). How would you dene the conditional densi-
ties of U
1
and U
2
given that U
1
> U
2
? What are E(U
1
| U
1
> U
2
) and
E(U
2
| U
1
> U
2
)?
13E-16 Suppose that the random variable B has the standard normal density.
What is the conditional probability density function of the sum of the two
roots of the quadratic equation x
2
+ 2Bx + 1 = 0 given that the two roots
are real?
13E-17 The transmission time of a message requires a geometrically dis-
tributed number of time slots, where the geometric distribution has param-
eter a with 0 < a < 1. In each time slot one new message arrives with
probability p and no message arrives with probability 1 p. What are the
expected value and the standard deviation of the number of newly arriving
messages during the transmission time of a message?
13E-18 In a buer there are a geometrically distributed number of messages
waiting to be transmitted over a communication channel, where the param-
eter p of the geometric distribution is known. Your message is one of the
waiting messages. The messages are transmitted one by one in a random
order. Let the random variable X be the number of messages that are trans-
mitted before your message. What are the probability mass function of X
and the expected value of X?
13E-19 The following game is oered in a particular carnival tent. The
carnival master has a red and a blue beaker each containing 10 balls numbered
as 1, . . . , 10. He shakes the beakers thoroughly and picks at random one ball
from each beaker. Then he tells you the value r of the ball picked from
the red beaker and asks you to guess whether the unknown value b of the
ball picked from the blue beaker is larger than r or smaller than r. If you
guesses correctly, you get b dollars. If r = b, you get
1
2
b dollars. If you are
wrong about which is larger, you get nothing. You have to pay $4.50 for the
23
privilege of playing the game. Is this a fair game?
6
13E-20 A bin contains N strings. You randomly choose two loose ends and
tie them up. You continue until there are no more free ends. What is the
expected number of loops you get?
13E-21 (a) You sample a random observation of a Poisson distributed random
variable with expected value 1. The result of the random draw determines
the number of times you toss a fair coin. What is the probability distribution
of the number of heads you will obtain?
(b) You perform the following experiment. First you generate a random
number from (0, 1). Then you simulate an integer by taking a random ob-
servation from a Poisson distribution whose expected value is given by the
random number you have generated. Let the random variable X be the in-
teger you will obtain from this experiment. What is the probability mass
function of X?
13E-22 You simulate a random observation from the random variable X
with the gamma density x
r1
[(1 p)/p]
r
e
(1p)x/p
/(r), where r and p are
given positive numbers. Then you generate an integer by taking a random
observation from a Poisson distribution whose expected value is given by the
number you have simulated from the gamma density. Let the random variable
N denote the generated integer. What is the probability mass function of
N?
13E-23 Let the random variable Z have the standard normal distribution.
Describe how to draw a random observation from|Z| by using the acceptance-
rejection method with the envelope density function g(x) = e
x
? How do
you get a random observation from Z when you have simulated a random
observation from |Z|?
13E-24 Your friend has fabricated a loaded die. In doing so, he has rst
simulated a number at random from the interval (0.1, 0.4). He tells you that
the die is loaded in such a way that any roll of the die results in the outcome
6 with a probability which is equal to the simulated number. Next the die is
rolled 300 times and you are informed that the outcome 6 has appeared 75
times. What is the posterior density of the probability that a single roll of
the die gives a 6?
13E-25 Your friend is a basketball player. To nd out how good he is in
free throws, you ask him to shoot 10 throws. Your priority density f
0
()
of the success probability of the free throws of your friend is a triangular
6
This problem is inspired by the Numberplay Blog of the New York Times, March 4,
2013.
24
density on (0.25, 0.75) with mode at 0.50. That is, f
0
() = 16( 0.25) for
0.25 < 0.50, f
0
() = 16(0.75 ) for 0.50 < < 0.75, and f
0
() = 0
otherwise. What is posterior density function f() of the success probability
given that your friend scores 7 times out of the 10 throws? What value of
maximizes the posterior density? Give a 95% Bayesian condence interval
for the success probability .
13E-26 A job applicant will have an IQ test. The prior density of his IQ is
the N(
0
,
2
0
) density with
0
= 100 and
0
= 15. If the true value of the
IQ of the job applicant is x, then the test will result in a score that has the
N(x,
2
1
) distribution with
1
= 7.5. The test results in a score of 123 points
for the job applicant. What is the posterior density f( | data) of the IQ
of the job applicant? Give the value of for which the posterior density is
maximal and give a 95% Bayesian condence interval for .
Chapter 14
14E-1 Let G
X
(z) be the generating function of the nonnegative, integer-
valued random variable X. Verify that P(X is even ) =
1
2
(G
X
(1) + 1).
14E-2 You rst roll a fair die once. Next you roll the die as many times
as the outcome of this rst roll. Let the random variable S be the sum of
the outcomes of all the rolls of the die, including the rst roll. What is the
generating function of S? What are the expected and standard deviation of
S?
14E-3 For any t 0, let N(t) be a Poisson distributed random variable with
expected value t for some > 0. Let the random variable T be independent
of the stochastic process {N(t), t 0} and have an exponential distribution
with parameter . What is the generating function of the random variable
X = N(T)? What do you conclude about the probability mass function of
X?
14E-4 You perform a sequence of independent Bernoulli trials with success
probability p. Let the random variable X be the number of trials needed
to obtain three successes in a row or three failures in a row. Determine the
generating function of X. Hint: let X
1
(X
2
) be the additional number of
trials to obtain three successes in a row or three failures in a row when the
rst trial results in a success (failure).
14E-5 Suppose that the random variable X has the so-called logistic density
function f(x) = e
x
/(1+e
x
)
2
for < x < . What is the interval on which
the moment-generating M
X
(t) is dened? Use M
X
(t) to nd the expected
value and variance of X.
25
14E-6 Suppose that the moment-generating function M
X
(t) of the continuous
random variable X has the property M
X
(t) = e
t
M
X
(t) for all t. What is
E(X)? Do you think that this property determines the density of X?
14E-7 You do not know the probability distribution of a random variable X,
but you do know its mean and standard deviation . Use Chebyshevs
inequality to answer the following two questions.
(a) What is a value for k such that the random variable X falls in the interval
( k, +k) with a probability of at least p?
(b) Prove that P(X > +a)

2
a
2
+
2
for any constant a > 0.
14E-8 (a) What is the moment-generating function of the random variable
X having the uniform distribution on the interval (1, 1)?
(b) Let X
1
, X
2
, . . . , X
n
be independent random variables that are uniformly
distributed on the interval (1, 1). Verify the Cherno bound P
_
1
n
(X
1
+
X
2
+ +X
n
) c
_
e

3
2
cn
2
for c > 0.
14E-9 Let X be a Poisson distributed random variable with expected value
. Give a Cherno bound for P(X c) when c > .
14E-10 In a branching process with one ancestor, the number of ospring of
each individual has the shifted geometric distribution {p
k
= p(1 p)
k
, k =
0, 1, . . .} with parameter p (0, 1). What is the probability of extinction as
function of p?
14E-11 At a production facility orders arrive one at a time, where the in-
terarrival times of the orders are independent random variables each having
the same distribution with expected value . A production is started only
when N orders have accumulated. The production time is negligible. A xed
cost of K > 0 is incurred for each production setup and holding costs are
incurred at the rate of hj when j orders are waiting to be processed. Identify
a regenerative stochastic process and nd the long-run average cost per unit
time as function of N. What value of N minimizes the long-run average cost
per unit time?
7
14E-12 Messages arriving at a communication channel according to a Poisson
process with rate are temporarily stored in a buer with ample capacity.
The buer is emptied every T time units, where T > 0 is xed. There is a
xed cost of K > 0 for emptying the buer. A holding cost at rate h > 0 per
unit time is incurred for each message in the buer. Identify a regenerative
stochastic process and nd the long-run average cost per unit time as function
of T. What value of T minimizes the long-run average cost per unit time?
7
The problems 14E-11 to 14E-15 are a bonus and illustrate the power of the strong law
of large numbers through the renewal-reward theorem.
26
14E-13 A canal touring boat departs for a tour through the canals of Ams-
terdam every T minutes with T xed. Potential customers pass the point of
departure according to a Poisson process with rate . A potential customer
who sees that the boat leaves s minutes from now joins the boat with prob-
ability e
s
for 0 s T. Assume that a xed cost of K > 0 is incurred for
each round trip and that a xed amount R > 0 is earned for each passen-
ger. Identify a regenerative stochastic process to nd the long-run average
net reward per unit time as function of T. What value of T maximizes this
long-run average net reward? Hint: use the fact the arrival time of the nth
potential customer has the density
n
t
n1
e
t
/(n 1)! and use conditioning
to nd the probability that the nth arrival joins the rst trip of the touring
boat.
14E-14 Customers asking for a single product arrive according to a Poisson
process with rate . Each customer asks for one unit of the product. Each
demand which cannot be satised directly from stock on hand is lost. Oppor-
tunities to replenish the inventory occur according to a Poisson process with
rate . This process is assumed to be independent of the demand process.
For technical reasons a replenishment can only be made when the inventory
is zero. The inventory on hand is raised to the level Q each time a replenish-
ment is done. Use the renewal reward theorem to nd the long-run fraction
of time the system is out of stock and the long-run fraction of demand that
is lost. Verify that these two fractions are equal to each other.
14E-15 Suppose that jobs arrive at a work station according to a Poisson
process with rate . The work station has no buer to store temporarily
arriving jobs. An arriving job is accepted only when the work station is
idle, and is lost otherwise. The processing times of the jobs are independent
random variables having a common probability distribution with nite mean
. Use the renewal-reward theorem to nd the long-run fraction of time
the work station is busy and the long-run fraction of jobs that are lost.
Conclude that these two fractions are equal to each other and that they use
the probability distribution of the processing time only through its expected
value.
27
Solutions of the Probability Exam Questions
Here we give the fully worked-out solutions of all of the probability exam
questions.
Chapter 7
7E-1 The choice of the sample space depends on whether we care about the
order in which the cards are chosen. If we consider the order in which the
13 cards are chosen as being relevant, we take an ordered sample space. Let
us assume that the 52 cards of the deck as 1, 2, . . . , 52 before the deck is
thoroughly shued. Imagine that the 13 cards of the bridge hand are chosen
one at a time. The ordered sample space is the set of outcomes all possible
ordered 13-tuples (i
1
, . . . , i
13
), where i
k
is the number of the kth chosen card.
The sample space has 52 51 40 equally likely outcomes. There are
32 31 20 outcomes for which there is no card above a nine among
the 13 cards chosen. Hence the probability of a Yarborough is
52 51 40
32 31 20
= 0.000547.
Alternatively, this probability can be computed by using an unordered sample
space. The order in which the 13 cards are chosen is not relevant in an
unordered sample space. Each outcome of this sample space is a set of 13
dierent cards from the deck of 52 cards. In a set we dont care which element
is rst, only which elements are actually present. The number of ways you
can choose a set of 13 dierent cards from a deck of 52 cards is given by the
binomial coecient
_
52
13
_
. Hence the unordered sample space has
_
52
13
_
equally
likely outcomes. The number of outcomes with no card above a nine is
_
32
13
_
.
This leads to the same value for the desired probability of a Yarborough:
_
32
13
_
_
52
13
_ = 0.000547.
This probability says that the odds against a Yarborough are 1827 to 1. A
Yarborough is named after Lord Yarborough (18091862). The second Earl
of Yarborough would oer his whist-playing friends a wager of 1,000 pounds
to 1 against them picking up such a hand. You see that the odds were on his
side. There is no record that he ever paid o.
7E-2 Take an ordered sample space whose outcomes are given by all possible
orderings of ve dierent cells. The total number of outcomes of the sample
space is 64 63 62 61 60 and all outcomes are equally probable. The
28
number of outcomes for which no row or column contains more than one dot
is 64 49 36 25 16. Hence the desired probability is
64 49 36 25 16
64 63 62 61 60
= 0.0494.
Alternatively, the desired probability can be computed by using an unordered
sample space. Take the sample space whose outcomes are given by all un-
ordered sets of ve dierent cells. The total number of outcomes of the
sample space is
_
64
5
_
. Noting that you can choose ve dierent rows in
_
8
5
_
ways, it follows that the number of choices of ve dierent cells for which no
row or column contains more than one dot is
_
8
5
_
8 7 6 5 4. This
gives the same result
_
8
5
_
8 7 6 5 4
_
64
5
_ = 0.0494
for the desired probability.
7E-3 Take an ordered sample space with as outcomes all possible 5! orderings
of the ve people, where the rst two people in the ordering get tea from the
waiter and the other three get coee. The number of orderings in which the
rst two people have ordered coee and the other people have ordered tea is
2! 3!. Hence the desired probability is
2!3!
5!
= 0.1. Alternatively, we can
take an unordered sample space whose outcomes are given by all possible
choices of two people from the ve people, where these two people get coee.
This leads to the same probability of 1/
_
5
2
_
= 0.1.
7E-4 Take an unordered sample space whose outcomes are given by all pos-
sible sets of the three dierent parking places out of the ten parking places.
The number of outcomes of the sample space is
_
10
3
_
. Each outcome of the
sample space gets assigned the same probability of 1/
_
10
3
_
. The number of
outcomes with three adjacent parking places is 8. Hence the desired proba-
bility is
8
_
10
3
_ =
1
15
.
7E-5 A same sample space can be used to answer both questions. Take
an ordered sample space whose outcomes are given by all possible 5-tuples
(i
1
, i
2
, . . . , i
5
), where i
k
is the exit oor for person k for k = 1, . . . , 5. The
sample space has 10
5
outcomes. Let us rst consider the case that the oors
are equally likely to be chosen as exit oor. Then the probability
1
10
5
is
assigned to each element of the sample outcomes. The number of outcomes
29
for which the ve persons are all going to dierent oors is 109876.
Hence the probability that they are all going to dierent oors is
10 9 8 7 6
10
5
= 0.3024.
For the case that the 10 oors are not equally likely to be chosen as the exit
oor, the probability
_
1
2
_
r
_
1
18
_
5r
is assigned to each outcome (i
1
, i
2
, . . . , i
5
)
for which exactly r of the components are equal to 10. The number of
outcomes for which the ve persons are all going to dierent oors and oor
10 is not chosen as the exit oor is equal to 9 8 7 6 5 = 15,120, while
the number of outcomes for which the ve persons are all going to dierent
oors and exactly one person chooses oor 10 as the exit oor is equal to
_
5
1
_
9 8 7 6 = 15,120. Then the probability that the persons are all
going to dierent oors is
15,120
_
1
18
_
5
+ 15,120
1
2
_
1
18
_
4
= 0.0800.
7E-6 Take as sample space the set of all possible orderings of the ten people
on the ten seats. The sample space has 10! equally likely outcomes.
(a) The number of ways to choose three adjacent seats from the ten seats is
8. Hence the number of outcomes for which the three friends are seated next
to each other is 8 3! 7! and so the probability that the three friends will
sit next to each other is
8 3! 7!
10!
=
1
15
.
(b) The number of ways to choose two friends from the three friends is
_
3
2
_
.
The number of outcomes for which the two chosen friends sit next to each
other at one of the ends of the row and the third friend has a non-adjacent
seat is 272! 7!, while the number of outcomes for which the two chosen
friends sit next to each other but not at one of the ends of the row and the
third friend has a non-adjacent seat is 7 6 2! 7!. Hence the probability
that exactly two of the three friends will sit next to each other is
_
3
2
__
2 7 2! 7! + 7 6 2! 7!

10!
=
7
15
.
7E-7 Take an ordered sample space whose outcomes are given by the 10!
possible orderings of the 10 people, where the rst ve people in the ordering
form the rst group and the other ve people form the second group. Each
element of the sample space is equally likely and so the desired probability
is
5 4 3 7! + 5 4 3 7!
10!
=
1
6
.
30
Alternatively, by taking an unordered sample space whose outcomes are given
by all possible sets of ve people for the rst group, the probability can be
computed as
_
7
2
_
+
_
7
5
_
_
10
5
_ =
1
6
.
7E-8 This problem is easiest solved by using an unordered sample space.
Take as sample space the set of all possible sets of four cards from the 52
cards. The sample space has
_
52
4
_
equally likely elements. The number sets
in which the four cards are J, Q, K, A is
_
4
1
_

_
4
1
_

_
4
1
_

_
4
1
_
. Hence the
desired probability is
_
4
1
_

_
4
1
_

_
4
1
_

_
4
1
_
_
52
4
_ = 9.46 10
4
.
You can also use an ordered sample space. Take an ordered sample space
whose outcomes are given by all 52! possible orderings of the 52 cards. There
are 4
4
4! 48! orderings for which the rst four cards are J, Q, K, A in any
order, with suit irrelevant. This gives the same probability
4
4
4! 48!
52!
= 9.46 10
4
.
7E-9 Label both the balls and the boxes as 1, 2, and 3. Take an ordered
sample space whose elements are all possible three-tuples (b
1
, b
2
, b
3
), where
b
i
is the label of the box in which ball i is dropped. The sample space has 3
3
equally likely elements. Let A
i
be the event that only box i is empty. The
events A
1
, A
2
and A
3
are mutually disjoint and have the same probability.
Hence the desired probability is P(A
1
A
2
A
3
) = 3P(A
1
). To nd P(A
1
),
note that he number of elements for which box 2 (3) contains two balls and
box 3 (2) contains one ball is
_
3
2
_
1 = 3. Hence the number of elements for
which only box 1 is empty is 2 3 = 6. This gives P(A
1
) =
6
27
and so the
probability that exactly one box will be empty is
2
3
.
7E-10 The sample space consists of all four-tuples (
1
,
2
,
3
,
4
) with
i
is
0 or 1, where
i
= 0 if component i has failed and
i
= 1 otherwise. The
probability r
1
r
2
r
3
r
4
is assigned to each outcome (
1
,
2
,
3
,
4
), where r
i
= f
i
if
i
= 0 and r
i
= 1f
i
if
i
= 1. It is easiest to compute the complementary
probability of no system failure. Let A be the event that none of the four
components has failed and B be the event that only one of the components 2,
3, or 4 has failed. Then P(AB) gives the probability of no system failure.
The events A and B are mutually disjoint and so P(AB) = P(A) +P(B).
Obviously, P(A) = (1f
1
)(1f
2
)(1f
3
)(1f
4
) and P(B) = (1f
1
)f
2
(1
31
f
3
)(1 f
4
) + (1 f
1
)(1 f
2
)f
3
(1 f
4
) + (1 f
1
)(1 f
2
)(1 f
3
)f
4
. Hence
the probability that the system will fail is given by
1
_
(1 f
1
)(1 f
2
)(1 f
3
)(1 f
4
) + (1 f
1
)f
2
(1 f
3
)(1 f
4
)
+(1 f
1
)(1 f
2
)f
3
(1 f
4
) + (1 f
1
)(1 f
2
)(1 f
3
)f
4

.
7E-11 The sample space of this experiment is the set {(x, y) : 0 x, y 1}.
The probability P(A) assigned to each subset A of the unit square is the area
of the set A. For xed a with 0 < a < 2, let A be the subset of points (x, y)
in the unit square that satisfy x + y a. The area of the set A is given by
1
2
a
2
for 0 < a 1 and by 1
1
2
(2a)
2
for 1 a < 2 (draw a picture). Note:
the probability of a Manhattan distance of no more than a is also equal to
1
2
a
2
for 0 < a 1 and to 1
1
2
(2 a)
2
for 1 a < 2 when the point is
randomly chosen in the square {(x, y) : 1 x, y 1}, as follows by using
a symmetry argument.
7E-12 The sample space is the set {(x, y) : 0 x 3, 0 y 2}. Let A be
the subset of points from the rectangle for which the distance to the closest
side of the rectangle is larger than a. Then A is a rectangle whose sides have
the lengths 32a and 22a (it is helpful to make a picture) and so the area
of A is (3 2a)(2 2a). It now follows that the desired probability is
6 (3 2a)(2 2a)
6
=
5
3
a
2
3
a
2
.
7E-13 For the case of n = 1, the sample space consists of the eight 3-tuples
(H, H, H), (H, H, T), (H, T, H), (H, T, T), (T, H, H), (T, H, T), (T, T, H),
and (T, T, T), where the rst two components refer to the outcomes of the
two coins of Pete. This gives that the probability of Pete getting more heads
than John is
4
8
=
1
2
for n = 1. By considering all possible 5-tuples of Hs and
Ts for the case of n = 2, the value
16
32
=
1
2
is found for the probability of
Pete getting more heads than John. For the general case the sample space
consists of the 2
2n+1
possible 2n + 1-tuples of Hs and Ts. The probability
of Pete getting more heads than John is then given by
n

k=0
_
n
k
_
n+1

j=k+1
_
n + 1
j
_
_
2
2n+1
.
Evaluating this expression for several values of n gives each time the value
1
2
. A very simple probabilistic argument can be given for the result that the
probability of Pete getting more heads than John is always equal to
1
2
. Let A
be the event that Pete gets more heads than John and B be the event that
32
Pete gets more tails than John. Since Pete has only coin more than John,
the events A and B are disjoint. In view of the fact that the total number of
coins is odd, it also holds that the union of A and B is the complete sample
space. Hence P(A B) = P(A) + P(B) = 1. By a symmetry argument
P(A) = P(B) and so P(A) =
1
2
as was to be veried.
7E-14 Take the same sample space as in the solution of Example 7.6 in the
book. Let A
i
be the event that the deadly shot falls at the ith trial. Then
P(A
i
) =
_
5
6
_
i1
1
6
for i = 1, 2, . . .. The events A
i
are mutually exclusive. The
probability that desperado A will shoot himself dead is
P
_

_
k=0
A
1+3k
_
=
1
6

k=0
_
5
6
_
3k
=
1/6
1 (5/6)
3
= 0.3956
The probability that desperado B will shoot himself dead is
P
_

_
k=0
A
2+3k
_
=
1
6

k=0
_
5
6
_
1+3k
=
5/36
1 (5/6)
3
= 0.3297.
The probability that desperado C will shoot himself dead is
P
_

_
k=0
A
3+3k
_
=
1
6

k=0
_
5
6
_
2+3k
=
25/216
1 (5/6)
3
= 0.2747.
7E-15 Let A be the event of getting three or more heads in a row in 20 tosses
of the coin and B be the event of getting three or more tails in a row in 20
tosses of the coin. Then P(A) = P(B) = 0.7870 and P(A B) = 0.9791.
Using the relation P(A B) = P(A) + P(B) P(AB), it follows that the
desired probability is given by
P(AB) = 0.7870 + 0.7870 0.9791 = 0.6049.
7E-16 Let A be the event that a second-hand car is bought and B be the
event that a Japanese car is bought. The desired probability is P(A B).
Denoting by E
c
the complement of even E, it follows that
P(A B) = P(A) +P(B) P(A B) = P(A) +P(B) [1 P((A B)
c
)]
= P(A) +P(B) [1 P(A
c
B
c
)]
= 0.25 + 0.30 [1 0.55] = 0.10.
7E-17 Let A
n
be the event that in the rst 6n rolls of the die each of the six
possible outcomes occurs n times. Then P(

n=1
A
n
) is an upper bound for
33
the probability that the six accumulating counts will ever be equal. For each
n,
P(A
n
) =
_
6n
n
__
5n
n
_

_
6n
n
_
6
6n
=
(6n)!
(n!)
6
6
6n
Using the fact that P(

n=1
A
n
)

n=1
P(A
n
), an upper bound for the
desired probability is given by

n=1
P(A
n
) = 0.02251,
where the innite sum has been numerically calculated.
7E-18 Take an ordered sample space whose outcomes are given by all 6
6
possible sequences of the integers 1, . . . , 6 to the length 6. All outcomes of
the sample space are equally likely. Let A
r
be the event that the largest
number rolled is r. Also, let B
k
the event that the largest number rolled is
smaller than or equal to k for k = 1, . . . , 6. Then B
r
= B
r1
A
r
. The
events B
r1
and A
r
are disjoint and so P(B
r
) = P(B
r1
) + P(A
r
). For any
r, P(B
r
) =
r
6
6
6
. Hence
P(A
r
) =
r
6
6
6

(r 1)
6
6
6
for r = 1, . . . , 6.
An alternative way to obtain P(A
r
) is as follows. For xed r, let E
j
be
the event that j of the six rolls of the die give the outcome r and the other
6 j rolls give an outcome less than r. Then A
r
=
6
j=1
E
j
. The events
E
1
, . . . , E
6
are mutually exclusive and so P(A
r
) =

6
j=1
P(E
j
). This gives
the alternative expression
P(A
r
) =
6

j=1
_
6
j
_
1
j
(r 1)
6j
6
6
.
The probability P(A
r
) has the numerical values 2.14 10
5
, 0.0014, 0.0143,
0.0722, 0.2471, and 0.6651 for r = 1, 2, 3, 4, 5, and 6.
7E-19 Some reection shows that the desired probability of Fermat being the
overall winner is the probability that Fermat wins at least 10 a games of
the additional 19 a b games. Take as sample space all possible sequences
of ones and zeros to a length of 19 a b, where a win of Fermat
is recorded as a one and a loss as a zero. Assign to each element of
the sample space the same probability of
1
2
19ab
. Let A
k
be the event that
34
Fermat wins exactly k of these games. The events A
k
are disjoint and so the
desired probability is
19ab

k=10a
P(A
k
) =
19ab

k=10a
_
1
2
_
19ab
.
7E-20 Let A
i
be the event that number i does not show up in any of the
50 drawings. By the inclusion-exclusion formula, the desired probability is
given by
P(A
1
A
2
A
10
) =
9

k=1
(1)
k+1
_
10
k
__
10 k
10
_
50
= 0.0509.
Note: The number of integers from 1, 2, . . . , 10 not showing up in 50 drawings
is approximately Poisson distributed with expected value = 10(9/10)
50
=
0.0515. This claim is supported by the fact that the Poisson probability
1 e

= 0.0502 is approximately equal to the probability 0.0509 that not


all of the numbers 1, 2, . . . , 10 will show up.
7E-21 Number the seats as 1, 2, . . . , 6 and take as sample space all 6! possible
ways the six people can be seated at the table. Let A
i
be the event that the
couple i sit together for i = 1, 2, 3. The desired probability is 1 P(A
1

A
2
A
3
). We have P(A
1
) = P(A
2
) = P(A
3
) = 6 2! 4!/6!, P(A
1
A
2
) =
P(A
1
A
3
) = P(A
2
A
3
) = 6 3 2! 2! 2!/6!, and P(A
1
A
2
A
3
) = 6 2 1
2! 2! 2!/6!. By the inclusion-exclusion formula,
P(A
1
A
2
A
3
) = 3P(A
1
)
_
3
2
_
P(A
1
A
2
) +P(A
1
A
2
A
3
) =
11
15
.
Hence the desired probability 1
11
15
=
4
15
.
7E-22 (a) Take as sample space all 6
6
possible sequences of the integers
1, . . . , 6 to the length of 6. The probability that three of the six possible
outcomes do not show up and each of the other three possible outcomes
shows up two times is
_
6
3
__
6
2
__
4
2
__
2
2
_
6
6
= 0.0386.
(b) Let A
i
be the event that outcome i shows up exactly three times for
i = 1, . . . , 6. Then the probability that some outcome shows up at least
three times is given by
P(A
1
A
6
) +
6

k=4
P(some outcome shows up exactly k times).
35
For any i, P(A
i
) =
_
6
3
_
5
3
/6
6
. Also, P(A
i
A
j
) =
_
6
3
_
/6
6
for any i, j with
i = j. By the inclusion-exclusion formula,
P(A
1
A
6
) = 6P(A
1
)
_
6
2
_
P(A
1
A
2
) = 0.3151.
The probability that any given outcome shows up exactly k times is equal to
_
6
k
_
5
6k
/6
6
for 4 k 6 and so
P(some outcome shows up exactly k times) = 6
6

k=4
_
6
k
_
5
6k
6
6
= 0.0522.
Hence the desired probability is 0.3151 + 0.0522 = 0.3672.
7E-23 Let A
i
be the event that the ith boy becomes part of a couple. The
desired probability is 1 P(A
1
A
n
). For any xed i,
P(A
i
) =
n n
2n2
n
2n
=
n
n
2
.
For any xed i and j with i = j,
P(A
i
A
j
) =
n (n 1)n
2n4
n
2n
=
n(n 1)
n
4
.
Continuing in this way, we nd
P(A
1
A
n
) =
n

k=1
(1)
k+1
_
n
k
_
n(n 1) (n k + 1)
n
2k
.
7E-24 There are four paths from node n
1
to node n
4
. These paths are the
paths (l
1
, l
5
), (l
2
, l
6
), (l
1
, l
3
, l
6
) and (l
2
, l
4
, l
5
). Let A
j
be the event that the jth
path is functioning. The desired probability is given by P(A
1
A
2
A
3
A
4
).
By the inclusion-exclusion formula,
P(A
1
A
2
A
3
A
4
) =
4

j=1
P(A
j
)
3

j=1
4

k=j+1
P(A
j
A
k
)
+
2

j=1
3

k=j+1
4

l=k+1
P(A
j
A
k
A
l
) P(A
1
A
2
A
3
A
4
).
Hence the probability that there is some functioning path from node n
1
to
node n
4
is equal to
p
1
p
5
+p
2
p
6
+p
1
p
3
p
6
+p
2
p
4
p
5
p
1
p
2
p
5
p
6
p
1
p
3
p
5
p
6
p
1
p
2
p
4
p
5
p
1
p
2
p
3
p
6
p
2
p
4
p
5
p
6
p
1
p
2
p
3
p
4
p
5
p
6
+p
1
p
2
p
3
p
5
p
6
+p
1
p
2
p
4
p
5
p
6
+p
1
p
2
p
3
p
4
p
5
p
6
+p
1
p
2
p
3
p
4
p
5
p
6
p
1
p
2
p
3
p
4
p
5
p
6
.
36
The four terms p
1
p
2
p
3
p
4
p
5
p
6
cancel out in this expression. For the special
case that p
i
= p for all i, the above expression simplies to
2p
2
(1 +p +p
3
) 5p
4
.
Chapter 8
8E-1 Take an ordered sample space whose outcomes are given by all possible
3-tuples (i
1
, i
2
, i
3
), where i
k
is the result of the roll of the kth die. Each of the
216 possible outcomes of the sample space gets assigned the same probability
of
1
216
. Let A be the event that the sum of the results of the rolls of the three
dice is 10 and B the event that the three dice show dierent results. The
desired probability is P(A | B) = P(AB)/P(B). The event AB consists of
the 3! orderings of {1, 3, 6}, the 3! orderings of {1, 4, 5} and the 3! orderings
of {2, 3, 5}. This gives P(AB) =
33!
216
=
18
216
. Clearly, P(B) =
654
216
=
120
216
.
8E-2 Take an ordered sample space whose outcomes are given by all 2
n
pos-
sible sequences of Hs and Ts to the length of n. Each outcome of the sample
space gets assigned the same probability of
1
2
n
. Let A be the event that
the rst toss gives heads and let B the event that r heads are obtained in
n tosses. Then P(B) =
_
n
r
_
/2
n
and P(AB) =
_
n1
r1
_
/2
n
. This leads to the
desired probability P(AB)/P(B) = r/n.
8E-3 The probability that the number of tens in the hand is the same as the
number of aces in the hand is given by
4

k=0
_
4
k
__
4
k
__
44
13 2k
_
_
_
52
13
_
= 0.3162.
Hence, using a symmetry argument, the probability that the hand contains
more aces than tens is
1
2
(10.3162) = 0.3419. By P(A | B) = P(AB)/P(B),
the other probability is equal to
0.341924

4
k=1
_
4
k
__
48
13k
_
/
_
52
13
_ = 0.4911.
8E-4 Let A be the event that a randomly chosen student chosen takes Spanish
and B be the event that the student takes French. The desired probability
is P(B | A) = P(AB)/P(A). It is given that P(A) = 0.35, P(B) = 0.15 and
P(A B) = 0.40. To nd P(AB), apply
P(A B) = P(A) +P(B) P(AB).
This gives P(AB) = 0.10 and so the desired probability is
0.10
0.35
=
2
7
.
37
8E-5 Since A = ABAB
c
and the events AB and AB
c
are mutually exclusive,
it follows that
P(A) = P(AB) +P(AB
c
) = P(A)P(B) +P(AB
c
),
by the independence of A and B. Hence
P(AB
c
) = P(A)[1 P(B)] = P(A)P(B
c
),
proving that A and B
c
are independent. By letting B
c
play the role of A
and A the role of B, it follows next that B
c
and A
c
are independent.
8E-6 The number is randomly chosen from the matrix and so P(A) =
30
50
,
P(B) =
25
50
and P(AB) =
15
50
. Since P(AB) = P(A)P(B), the events A and
B are independent. This result can also be explained by noting that you
obtain a random number from the matrix by choosing rst a row at random
and choosing next a column at random.
8E-7 (a) Imagining that the waiter rst serves the two cups of coee, let
A
1
be the event that the rst coee is given to a person having ordered
coee and A
2
be the event that the second coee is given to the other person
having ordered coee. The desired probability P(A
1
A
2
) can be evaluated as
P(A
1
)P(A
2
| A
1
) =
2
5

1
4
=
1
10
.
(b) Let A
1
be the event that your rst friend is in the same group as you are
and A
2
the event that your second friend is in the same group as you are.
The desired probability is P(A
1
A
2
) = P(A
1
)P(A
2
| A
1
) =
4
9

3
8
=
1
6
.
(c) Let A
i
be the event that the ith card you receive is a picture card that you
have not received before. Then, by P(A
1
A
2
A
3
A
4
) = P(A
1
)P(A
2
| A
1
)P(A
3
|
A
1
A
2
)P(A
4
| A
1
A
2
A
3
), the desired probability can also be computed as
P(A
1
A
2
A
3
A
4
) =
16
52

12
51

8
50

4
49
= 9.46 10
4
.
8
8E-8 Let A
i
be the event that one red and one blue ball show up on the ith
drawing for i = 1, . . . , 4. The desired probability is given by
P(A
1
A
2
A
3
A
4
) = P(A
1
)P(A
2
| A
1
)P(A
3
| A
1
A
2
)P(A
4
| A
1
A
2
A
3
).
We have P(A
1
) =
84
87
=
4
7
, P(A
2
| A
1
) =
63
65
=
3
5
, P(A
3
| A
1
A
2
) =
42
43
=
2
3
,
and P(A
4
| A
1
A
2
A
3
) = 1 (alternatively, P(A
1
) =
_
4
1
__
4
1
_
/
_
8
2
_
, and so on).
Hence the desired probability is given by
4
7

3
5

2
3
1 =
8
35
.
8
Problem 8E-4 illustrates again that conditioning is often easier than counting.
38
8E-9 For xed n, let A
n
be the event that your name appears at the nth
drawing and A
i
be the event that your name does not appear at the ith
drawing for i = 1, . . . , n 1. Using the basic formula P(A
1
A
2
A
n
) =
P(A
1
)P(A
2
| A
1
) P(A
n
| A
1
. . . A
n1
), it follows that
P(you get assigned the nth placement) =
n1

k=1
N k
2 +N k

2
2 +N n
.
8E-10 Let p
k
be the probability that the kth announcement wins the bonus.
Then, by P(A
1
A
2
A
n
) = P(A
1
)P(A
2
| A
1
) P(A
n
| A
1
. . . A
n1
), it
follows that p
1
= 0, p
2
=
1
100
, p
3
=
99
100

2
100
, and
p
k
=
99
100

98
100

99 k + 3
100

k 1
100
, k = 4, . . . , 25.
Numerical calculations show that p
k
is maximal for k = 11 with p
11
= 0.0628.
8E-11 Let p
k
be the probability that the rst ace appears at the kth card.
Then, by P(A
1
A
2
A
n
) = P(A
1
)P(A
2
| A
1
) P(A
n
| A
1
. . . A
n1
), it
follows that p
1
=
4
52
, p
2
=
48
52

4
51
, and
p
k
=
48
52

47
51

48 k + 2
52 k + 2

4
52 k + 1
, k = 3, . . . , 49.
The probability that player A becomes the dealer is

16
n=0
p
1+3n
= 0.3600,
the probability that player B becomes the dealer is

15
n=0
p
2+3n
= 0.3328, and
the probability that player C becomes the dealer is

15
n=0
p
3+3n
= 0.3072.
8E-12 Let A be the event that HAPPY HOUR appears again. To nd P(A),
condition on the events B
1
, B
2
, and B
3
, where B
1
is the event that the two
letters H are removed, B
2
is the event that the two letters P are removed,
and B
3
is the event that two dierent letters are removed. The formula
P(A) =

3
i=1
P(A | B
i
)P(B
i
) gives
P(A) = 1
1
_
9
2
_ + 1
1
_
9
2
_ +
1
2

_
1
1
_
9
2
_
1
_
9
2
_
_
=
19
36
.
8E-13 Let A be the event that the rst question on the exam is marked as
dicult. Let B
1
be the event that the exam is hard and B
2
be the event that
the exam is easy. Applying the formula P(A) =

2
i=1
P(A | B
i
)P(B
i
) gives
P(A) = 0.9 0.8 + 0.15 0.1 = 0.735.
39
The probability that the exam is hard given that the rst question on the
exam is marked as dicult is equal to
P(B
1
| A) =
P(AB
1
)
P(A)
=
P(A | B
1
)P(B
1
)
P(A)
=
0.9 0.8
0.735
= 0.9796.
8E-14 (a) Let A
k
be the event that both players have obtained k heads after
N tosses. The probability of a tie after N tosses is
N

k=0
P(A
k
) =
N

k=0
_
N
k
__
1
2
_
N
_
N
k
__
1
2
_
N
=
_
1
2
_
2N N

k=0
_
N
k
__
N
N k
_
=
_
1
2
_
2N
_
2N
N
_
.
Note that this probability is the same as the probability of getting N heads
(and N tails) in 2N tosses of a fair coin.
(b) Let A be the event that player 1 wins the game. Then P(A) = 0.5,
regardless of the value of N. The simplest way to see this is to dene E
1
as
the event that player 1 has more heads than player 2 after N tosses, E
2
as
the event that player 1 has fewer heads than player 2 after N tosses, and E
3
as the event that player 1 has the same number of heads as player 2 after
N tosses. Then P(A) =

3
i=1
P(A | E
i
)P(E
i
), by the law of conditional
probability. To evaluate this, it is not necessary to know the P(E
i
). Since
P(E
2
) = P(E
1
) and P(E
3
) = 1 2P(E
1
), it follows that P(A) is equal to
1 P(E
1
) + 0 P(E
2
) +
1
2
P(E
3
) = P(E
1
) +
1
2

_
1 2P(E
1
)
_
= 0.5.
8E-15 Let A be the event that all of the balls drawn are blue and B
i
be the
event that the number of points shown by the die is i for i = 1, . . . , 6.
(a) By the law of conditional probability, the probability that all of the balls
drawn are blue is given by
P(A) =
6

i=1
P(A | B
i
)P(B
i
) =
1
6
5

i=1
_
5
i
_
_
10
i
_ =
5
36
.
(b) The probability that the number of points shown by the die is r given
that all of the balls drawn are blue is equal to
P(B
r
| A) =
P(B
r
A)
P(A)
=
(1/6)
_
5
r
_
/
_
10
r
_
5/36
.
40
This probability has the values
3
5
,
4
15
,
1
10
,
1
35
,
1
210
and 0 for r = 1, . . . , 6.
8E-16 Let A be the event that John and Pete meet each other in the semi-
nals. To nd P(A), let B
1
be the event that John and Pete are allocated
to either group 1 or group 2 but are not in the same group and B
2
be the
event that John and Pete are allocated to either group 3 or group 4 but are
not in the same group. Then P(B
1
) = P(B
2
) =
1
2

2
7
=
1
7
. By the law of
conditional probability,
P(A) = P(A | B
1
)
1
7
+P(A | B
2
)
1
7
=
1
2

1
2

1
7
+
1
2

1
2

1
7
=
1
14
.
Let C be the event that John and Pete meet each other in the nal. To nd
P(C), let D
1
be the event that John is allocated to either group 1 or group
2 and Pete to either group 3 or group 4 and D
2
be the event that John is
allocated to either group 3 or group 4 and Pete to either group 1 or group 2.
Then P(D
1
) = P(D
2
) =
1
2

4
7
=
2
7
. By the law of conditional probability,
P(C) = P(C | D
1
)
2
7
+P(C | D
2
)
2
7
=
1
2

1
2

1
2

1
2

2
7
+
1
2

1
2

1
2

1
2

2
7
=
1
28
.
The latter result can also be directly seen by a symmetry argument. The
probability that any one pair contests the nal is the same as that for any
other pair. There are
_
8
2
_
dierent pairs and so the probability that John and
Pete meet each other in the nal is 1/
_
8
2
_
=
1
28
.
8E-17 Let P(r, s) be the probability of Fermat being the overall winner given
that Fermat has won so far r games and Pascal has won so far s games. Then,
by conditioning on the result of the next game,
P(r, s) =
1
2
P(r + 1, s) +
1
2
P(r, s + 1)
with the boundary conditions P(10, s) = 1 for all s < 10 and P(r, 10) = 0 for
all r < 10. By backward calculation, we get the desired probability P(a, b).
8E-18 (a) For xed n, let A
n
be the event that the total number of heads
after n tosses is even and let P
n
= P(A
n
). Also, let B
0
be the event that
the rst toss results in tails and B
1
be the event that the rst toss results in
heads. Then, P(A
n
) = P(A
n
| B
0
)P(B
0
) +P(A
n
| B
1
)P(B
1
). This gives
P
n
= (1 p)P
n1
+p(1 P
n1
) for n = 1, 2, . . . ,
41
where P
0
= 1. This recurrence equation has the explicit solution
P
n
=
1
2
+
1
2
(1 2p)
n
for n = 1, 2, . . . .
Note that, for any 0 < p < 1, the probability P
n
tends to
1
2
as n gets large.
For p =
1
2
, P
n
=
1
2
for all n 1, as is obvious for reasons of symmetry.
(b) For xed n, let A
n
be the event that a sequence of n tosses does not
show ve or more consecutive heads and let P
n
= P(A
n
). Also, let B
i
be
the event that the rst i 1 tosses give heads and the ith toss gives tails
for i = 1, 2, . . . , 5 and B
6
be the event that the rst ve tosses result in
heads. The events B
i
are disjoint. Then P(A
n
) =

6
i=1
P(A
n
| B
i
)P(B
i
)
with P(B
i
) = p
i1
(1 p) for 1 i 5 and P(A
n
| B
6
) = 0. This gives the
recursion
P
n
=
5

i=1
p
i1
(1 p)P
ni
for n = 5, 6, . . . ,
where P
0
= P
1
= P
2
= P
3
= P
4
= 1.
8E-19 For xed integer r, let A
r
be the event that there are exactly r winning
tickets among the one half million tickets sold. Let B
n
be the event that there
exactly n winning tickets among the one million tickets printed. Then, by
the law of conditional probability,
P(A
r
) =

n=0
P(A
r
| B
n
)P(B
n
).
Obviously, P(A
r
| B
n
) = 0 for n < r. For all practical purposes the so-called
Poisson probability e
1
/n! can be taken for the probability of the event B
n
for n = 0, 1, . . . (see Example 7.14 in the textbook). This gives
P(A
r
) =

n=r
_
n
r
__
1
2
_
r
e
1
n!
= e
1
(1/2)
r
r!

j=0
1
j!
= e

1
2
(1/2)
r
r!
.
Hence the probability of exactly r winning tickets among the one half million
tickets sold is given by the Poisson probability e

1
2
(1/2)
r
r!
for r = 0, 1, . . ..
8E-20 Let the hypothesis H be the event that a 1 is sent and the evidence E
be the event that a 1 is received. The desired posterior probability P(H | E)
satises
P(H | E)
P(H | E)
=
0.8
0.2

0.95
0.01
= 380.
Hence the posterior probability P(H | E) that a 1 has been sent is
1
1+380
=
0.9974.
42
8E-21 Let the hypothesis H be the event that the inhabitant you overheard
spoke truthfully and the evidence E be the event that the other inhabitant
says that the inhabitant you overheard spoke the truth. The desired posterior
probability P(H | E) satises
P(H | E)
P(H | E)
=
1/3
2/3

1/3
2/3
=
1
4
.
Hence the posterior probability P(H | E) that the inhabitant you overheard
spoke the truth is
1/4
1+1/4
=
1
5
.
8E-22 Let the hypothesis H be the event that oil is present in the test area
and the evidence E be the event that the test gives a positive signal. This
leads to
P(H | E)
P(H | E)
=
0.4
0.6

0.90
0.15
= 4.
Hence the posterior probability P(H | E) of oil being present in the test area
is
4
1+4
= 0.8.
8E-23 Let the hypotheses H be the event that the family has two girls and
the evidence E be the event that the family has a girl named Mary Ann. In
the rst problem, the prior probabilities P(H) and P(H) have the values
1
4
and
3
4
, while P(E | H) and P(E | H) have the values 1 and
2
3
. This leads to
P(H | E)
P(H | E)
=
1/4
3/4

1
2/3
=
1
2
and so the posterior probability P(H | E) of having two girls has the value
1
3
in the rst problem. In the second problem, the prior probabilities P(H)
and P(H) have again the values
1
4
and
3
4
, but P(E | H) and P(E | H) now
have the values
1
10
+
9
10

1
9
=
2
10
and
2
3

1
10
. This gives
P(H | E)
P(H | E)
=
1/4
3/4

2/10
(2/3) (1/10)
= 1
and so the posterior probability P(H | E) of having two girls has the value
1
2
in the second problem. An intuitive explanation of why the second proba-
bility is larger lies in the fact that in the second problem it is more likely to
have a girl named Mary Ann when there are two girls in the family rather
than a single girl. In the rst problem it makes no dierence whether the
number of girls in the family is one or two in order to have a girl named
Mary Ann.
43
8E-24 Let the hypothesis H be the event that the family has one boy and
two girls and the evidence E be the event that he family has a boy among
the three children. Then P(H) =
3
8
, P(H) =
5
8
, P(E | H) = 1, and P(E |
H) =
4
5
P(H | E)
P(H | E)
=
3/8
5/8

1
4/5
=
3
4
.
Hence the desired probability P(H | E) is equal to
3/4
1+3/4
=
3
7
.
8E-25 (a) Assume you have tossed the randomly chosen coin n times and all
n tosses result in heads. Let the hypothesis H be the event that you have
chosen the two-headed coin and the evidence E be the event that all n tosses
result in heads. The desired posterior probability P(H | E) satises
P(H | E)
P(H | E)
=
1/10,000
9,999/10,000

1
0.5
n
.
This gives
P(H | E) =
2
n
2
n
+ 9,999
.
The posterior probability P(H | E) has the values 0.7662 and 0.9997 for
n = 15 and n = 25.
(b) Let hypothesis H be the event that you choose coin s and the evidence
E be the event that each of the rst n tosses results in heads. By Bayes rule
in odds form,
P(H | E)
P(H | E)
=
1/(r + 1)
r/(r + 1)

(s/r)
n
(1/r)

j=s
(j/r)
n
=
(s/r)
n

j=s
(j/r)
n
.
This gives
P(H | E) =
(s/r)
n

r
j=0
(j/r)
n
.
8E-26 (a) Let the hypothesis H be the event that both numbers are even and
the evidence E be the event that there is an even number among the two
numbers. Then
P(H | E)
P(H | E)
=
25/100
75/100

1
(2 5 5)/100
=
2
3
.
Hence the posterior probability P(H | E) that both numbers are even is
2/3
1+2/3
= 0.4.
(b) Let the hypothesis H be the event that both numbers are even and the
44
evidence E be the event that there the number 2 is among the two numbers.
Then
P(H | E)
P(H | E)
=
25/100
75/100

1 (4/5)
2
(2 5)/100
=
6
5
.
Hence the posterior probability P(H | E) that both numbers are even is
6/5
1+6/5
=
6
11
.
8E-27 Let the random variable represent the unknown probability that a
single toss of the die results in the outcome 6. The prior distribution of is
given by p
0
() = 0.25 for = 0.1, 0.2, 0.3 and 0.4. The posterior probability
p( | data) = P( = | data) is proportional to L(data | )p
0
(), where
L(data | ) =
_
300
75
_

75
(1 )
225
. Hence the posterior probability p( | data)
is given by
p( | data) =
L(data | )p
0
()

4
k=1
L(data | k/10)p
0
(k/10)
=

75
(1 )
225

4
k=1
(k/10)
75
(1 k/10)
225
, = 0.1, 0.2, 0.3, 0.4.
The posterior probability p( | data) has the values 3.5 10
12
, 0.4097,
0.5903, and 3.5 10
12
for = 0.1, 0.2, 0.3, and 0.4.
8E-28 Let the random variable represent the unknown probability that
a free throw of your friend will be successful. The prior probability mass
function p
0
() = P( = ) has the values 0.2, 0.6, and 0.2 for = 0.25,
0.50, and 0.75. The posterior probability p( | data) = P( = | data) is
proportional to L(data | )p
0
(), where L(data | ) =
_
10
7
_

7
(1 )
3
. Hence
the posterior probability p( | data) is given by

7
(1 )
3
p
0
()
0.25
7
0.75
3
0.2 + 0.50
7
0.50
3
0.6 + 0.75
7
0.25
3
0.2
for = 0.25, 0.50, and 0.75. The possible values 0.25, 0.50 and 0.75 for
the success probability of the free throws of your friend have the posterior
probabilities 0.0051, 0.5812 and 0.4137.
Chapter 9
9E-1 This problem can be solved both by counting arguments and by con-
ditional probabilities. The solution approach using counting arguments re-
quires the specication of a sample space. Take as sample space the set of all
possible 10! rankings of the scores of the ten people. The number of rankings
45
for which the highest ranking achieved by a women equals 1 is given by 59!.
Hence
P(X = 1) =
5 9!
10!
=
5
10
.
The number of rankings for which the highest ranking achieved by a women
equals 1 is given by 5 4 8!. Hence
P(X = 2) =
5 4 8!
10!
=
5
18
.
Continuing in this way,
P(X = 3) =
5 4 5 7!
10!
=
5
36
, P(X = 4) =
5 4 3 5 6!
10!
=
5
84
,
P(X = 5) =
5 4 3 2 5 5!
10!
=
5
252
,
P(X = 6) =
5 4 3 2 1 5 4!
10!
=
1
252
.
To solve the problem by conditional probabilities, let us note that the problem
is one of the many problems that can be reformulated as balls in a bag
problem. Imagine that there are ve red balls and ve blue balls in a bag.
You pull them out one at a time. Let A
i
be the event that the ith ball pulled
out is the rst blue ball you get. Then P(X = i) = P(A
i
) for i = 1, . . . , 6.
Using the relation
P(A
i
) = P(A
c
1
) P(A
c
i1
| A
c
1
A
c
i2
)P(A
i
| A
c
1
A
c
i1
),
where A
c
i
is the complement event of A
i
, it follows that
P(X = 1) =
5
10
, P(X = 2) =
5
10

5
9
, P(X = 3) =
5
10

4
9

5
8
,
P(X = 4) =
5
10

4
9

3
8

5
7
, P(X = 5) =
5
10

4
9

3
8

2
7

5
6
,
P(X = 6) =
5
10

4
9

3
8

2
7

1
6

5
5
.
Alternatively, P(A
i
) can be calculated as
(
5
i1
)
(
10
i1
)

5
10(i1)
for i = 1, . . . , 6.
9E-2 The random variable X is dened on a sample space consisting of the
36 equiprobable elements (i, j), where i, j = 1, 2, . . . , 6. The random variable
X takes on the value i j for the realization (i, j). The random variable X
takes on the 1, 2, 3, 4, 5, 6, 8, 9, 10, 12, 15, 16, 18, 20, 24, 25, 30, and 36
46
with respective probabilities
1
36
,
2
36
,
2
36
,
4
36
,
2
36
,
4
36
,
2
36
,
1
36
,
2
36
,
4
36
,
2
36
,
1
36
,
2
36
,
2
36
,
2
36
,
1
36
, and
1
36
. The expected value of X is easiest computed as
E(X) =
1
36
6

i=1
6

j=1
i j =
1
6
6

i=1
i
1
6
6

j=1
j =
7
2

7
2
=
49
4
.
Also,
E(X
2
) =
1
36
6

i=1
6

j=1
(i j)
2
=
1
6
6

i=1
i
2

1
6
6

j=1
j
2
=
91
6

91
6
=
8281
36
.
Hence E(X) = 12.25 and (X) =
_
8281/36 (49/4)
2
= 8.9423.
9E-3 Let the random variable X be the payo (expressed in dollars) on a
lottery ticket. The random variable X takes on the values 0, 50, 500, and
5,000 with respective probabilities
P(X = 0) =
1 9 10 10 + 9 10 10 10
10,000
=
99
100
,
P(X = 50) =
1 1 9 10
10,000
=
9
1,000
,
P(X = 500) =
1 1 1 9
10, 000
=
9
10,000
, P(X = 5,000) =
1
10,000
.
This gives
E(X) = 0
99
100
+ 50
9
1,000
+ 500
9
10,000
+ 5,000
1
10, ,000
= 1.4.
The house edge of the lottery is
_
21.4
2
_
100% = 30%.
9E-4 Let the random variable X be your payo expressed in dollars. The
random variable X can take the values 0.5, 5, 6, 7, 8, 9, and 10 with respective
probabilities
4
10
,
1
10
,
1
10
,
1
10
,
1
10
,
1
10
, and
1
10
. This gives
E(X) = 0.50
4
10
+
10

j=5
j
1
10
= 4.7.
To calculate (X), note that
E(X
2
) = 0.50
2

4
10
+
10

j=5
j
2

1
10
= 35.6.
47
Hence (X) =

35.6 4.7
2
= 3.68.
9E-5 Let the random variable X be the smallest number rolled. The easiest
way to calculate E(X) is to use E(X) =

5
j=0
P(X > j). Noting that
P(X > j) =
_
6 j
6
_
6
for j = 0, 1, . . . , 5,
it follows that
E(X) =
5

j=0
_
6 j
6
_
6
= 1.4397.
To nd (X), note that P(X = k) = P(X > k 1) P(X > k) and so
E(X
2
) =
6

k=1
k
2
_
_
6 k + 1
6
_
6

_
6 k
6
_
6
_
= 2.5656.
This gives (X) =
_
E(X
2
) E
2
(X) = 0.702.
9E-6 The idea is to use a one-stage-look-ahead rule. For the situation that
you have collected so far a dollars and k + 1 boxes are still closed including
the box with the devils penny, dene the random variable X
k
(a) as the
amount by which your capital would change when you would decide to open
one other box. The one-stage-look-ahead-rule prescribes to open one other
box if the expected value of X
k
(a) is positive and to stop otherwise. Let
A = a
1
+ a
2
+ + a
10
be the original amount of dollars in the 10 boxes.
To calculate E
_
X
k
(a)
_
, you need not to know how the remaining amount of
Aa dollars is distributed over the k remaining closed boxes not containing
the devils penny. To see this, imagine that the dollar amounts a
i
1
, . . . , a
i
k
are in these k closed boxes. Then,
E
_
X
k
(a)
_
=
1
k + 1
k

j=1
a
i
j

1
k + 1
a =
1
k + 1
(A a)
1
k + 1
a.
Hence E
_
X
k
(a)
_
> 0 only if a <
1
2
A, regardless of the value of k. In other
words, a reasonable strategy is to stop as soon you have collected
1
2
A dollars
or more and to continue otherwise.
Remark. The following game is a special case of Problem 9E-6. A game
is played with eleven cards: an ace, two, three, . . ., nine, ten, and a joker.
These cards are thoroughly shued and laid face down. You can ip over as
many cards as you wish, but they must be ipped over one by one. Once you
ip over the joker card, the game is over and you lose all the money gathered
48
so far. If you stop before the joker card is ipped over, you win as many
dollars as the sum of the values of the cards you have ipped over.
9E-7 For any k 2, let X
k
be the amount you get at the kth game. Then
the total amount you will get is

s
k=2
X
k
. By the linearity of the expectation
operator, the expected value of the total amount you will get is

s
k=2
E(X
k
).
Since P(X
k
= 1) =
1
k(k1)
and P(X
k
= 0) = 1 P(X
k
= 1), it follows that
E(X
k
) =
1
k(k1)
for k = 2, . . . , s. Hence the expected value of the total
amount you will get is equal to
s

k=2
1
k(k 1)
=
s 1
s
.
The fact that the sum equals
s1
s
is easily proved by induction.
9E-8 By conditioning, the probability of getting the same two outcomes when
tossing the two coins is
1
2
p +
1
2
(1 p) =
1
2
. Hence the number of tosses
until both coins simultaneously show the same outcome is geometrically dis-
tributed with parameter
1
2
and thus has expected value 2 and standard de-
viation

2.
9E-9 For xed k, let the random variable X
i
be equal to 1 if box i contains
exactly k apples and X
i
be equal to 0 otherwise. Then,
P(X
i
= 1) =
_
25
k
__
1
10
_
k
_
9
10
_
25k
.
Hence the expected value of the number of boxes containing exactly k apples
is given by
E(X
1
+ +X
10
) = 10
_
25
k
__
1
10
_
k
_
9
10
_
25k
, k = 0, 1, . . . , 25.
9E-10 Let us say that a success occurs each time an ace is drawn that you
have not seen before. Denote by X
j
be the number of cards drawn between
the occurrences of the (j 1)st and jth success. The random variable X
j
is geometrically distributed with success parameter
4(j1)
52
. Also, the ran-
dom variables X
1
, . . . , X
4
are independent of each other. A geometrically
distributed random variable with parameter p has expected value 1/p and
variance (1 p)/p
2
. Hence the expected value and the standard deviation
of the number of times you have to draw a card until you have seen all four
dierent aces are given by
E(X
1
+X
2
+X
3
+X
4
) =
52
4
+
52
3
+
52
2
+
52
1
= 108.33
49
and
(X
1
+X
2
+X
3
+X
4
) =

_
4

k=1
1 k/52
(k/52)
2
= 61.16.
9E-11 Let the random variable X
j
= 1 if the elevator stops on oor j and
X
j
= 0 otherwise. It holds that P(X
j
= 0) =
_
r1
r
_
m
and so E(X
j
) =
1
_
r1
r
_
m
for all for j = 1, 2, . . . , r. Hence the expected value of the number
of stops of the elevator is equal to
E(X
1
+X
2
+ +X
r
) = r
_
1
_
r 1
r
_
m
_
.
The variance of the number of stops of the elevator is given by

2
(X
1
+X
2
+ +X
r
) =
r

i=1
E(X
2
i
) + 2
r

i=1
r

j=i+1
E(X
i
X
j
).
Since X
i
is a 0 1 variable, we have E(X
2
i
) = E(X
i
) = 1
_
r1
r
_
m
for all
for i. To nd E(X
i
X
j
) for i = j, we use the relation
P(X
i
= 1, X
j
= 1) = 1 [P(X
i
= 0) +P(X
j
= 0) P(X
i
= 0, X
j
= 0)].
For any i = j,
P(X
i
= 0, X
j
= 0) =
_
r 2
r
_
m
.
Putting the pieces together, we nd that the standard deviation of the num-
ber of stops of the elevator is given by

+ 2
_
r
2
_
_
1 2
_
r 1
r
_
m
+
_
r 2
r
_
m _

2
,
where = r
_
1
_
r1
r
_
m
_
.
9E-12 (a) Let X
i
be equal to 1 if the integer i does not show up in the 20
drawings and X
i
be equal to 0 otherwise. Then the number of integers not
showing in the 20 drawings is X
1
+ X
2
+ + X
10
. For all i, P(X
i
= 1) =
(9/10)
20
and so
E(X
1
+X
2
+ +X
10
) = 10
_
9
10
_
20
= 1.2158.
50
For any i = j, P(X
i
= 1, X
j
= 1) = (8/10)
20
and so
E(X
1
+X
2
+ +X
10
)
2
= 10
_
9
10
_
20
+ 2
_
10
2
_

_
8
10
_
20
= 2.2534.
This leads to (X
1
+ X
2
+ + X
10
) = 0.8805. Note: The Problems 9E-6
and 9E-7(a) are instances of the same model.
(b) Let X
i
be equal to 1 if the number i does not show up in the 15 lotto
drawings and X
i
be equal to 0 otherwise. The probability that a specied
number does not show up in any given drawing is
_
44
6
_
/
_
45
6
_
= 39/45. Hence
P(X
i
= 1) = (39/45)
15
and so
E(X
1
+X
2
+ +X
45
) = 45
_
39
45
_
15
= 5.2601.
The probability that two specied numbers i and j with i = j do not show
up in any given drawing is
_
43
6
_
/
_
45
6
_
= (39/45) (38/44). Hence P(X
i
=
1, X
j
= 1) = [(39 38)/(45 44)]
20
and so
E(X
1
+X
2
+ +X
45
)
2
= 45
_
39
45
_
15
+ 2
_
45
2
_

_
39 38
45 44
_
15
= 30.9292.
This leads to (X
1
+X
2
+ +X
45
) = 1.8057.
9E-13 For any i = j, let the random variable X
ij
= 1 if the integers i and j
are switched in the random permutation and X
ij
= 0 otherwise. The total
number of switches is

i<j
X
ij
. Since P(X
ij
= 1) =
(n2)!
n!
, it follows that
E(X
ij
) =
1
n(n1)
. Hence
E
_

i<j
X
ij
_
=

i<j
E(X
ij
) =
_
n
2
_
1
n(n 1)
=
1
2
,
irrespective of the value of n.
9E-14 Call your opponents East and West. The probability that East has
two spades and West has three spades is
_
5
2
__
21
10
_
_
26
13
_ =
39
115
.
Hence the desired probability is 2
39
115
= 0.6783.
51
9E-15 Let the random variable X denote the number of times you will roll
the die until you get an outcome that is larger than or equal to the randomly
chosen integer from 1, 2, . . . , 6. Under the condition that the randomly chosen
integer is j, the number of rolls of the die until you get an outcome that is
larger than or equal to j is geometrically distributed with parameter p
j
=
6j+1
6
. By the law of conditional probability, it now follows that
P(X = k) =
6

j=1
_
1
6 j + 1
6
_
k1
6 j + 1
6

1
6
, k = 1, 2, . . . .
Using the fact that the geometric distribution with parameter p has 1/p as
expected value and (1 p)/p
2
as variance, it follows that
E(X) =
1
6

k=1
k
6

j=1
_
1
6 j + 1
6
_
k1
6 j + 1
6
=
1
6
6

j=1
6
6 j + 1
,
yielding E(X) = 2.45. Also,
E(X
2
) =
1
6

k=1
k
2
6

j=1
_
1
6 j + 1
6
_
k1
6 j + 1
6
=
6

j=1
6 +j 1
(6 j + 1)
2
,
yielding E(X
2
) = 15.4467. This gives (X) =

15.4467 2.45
2
= 3.073.
9E-16 Let the indicator variable I
n
= 1 if the nth toss results in heads and
I
n
= 0 otherwise. To nd E(

r
n=1
I
n
), let a
n
= P(I
n
= 1) denote the
probability that the nth toss will result in heads. By the law of conditional
probabilities, P(I
n
= 1) is given by P(I
n
= 1 | I
n1
= 1)P(I
n1
= 1) +
P(I
n
= 1 | I
n1
= 0)P(I
n1
= 0) and so
a
n
= 0.5a
n1
+p(1 a
n1
) for n = 0, 1, . . . ,
where a
0
= 1. The solution of this standard dierence equation is given by
a
n
= +
n
for n = 1, 2, . . . ,
where =
p
p+0.5
, =
1
2p+1
and = 0.5 p. This gives that the expected
value of the number of heads in r tosses is given by

r
n=1
( +
n
).
9E-17 (a) You can think of n + m independent Bernoulli experiments with
success probability p, where X is the number of successes in the rst n
experiments and Y is the number of successes in the last m experiments.
This explains why X + Y is binomially distributed with parameters n + m
52
and p. A formal proof goes as follows. Using the independence of X and Y ,
it follows that
P(X +Y = k) =
k

r=0
P(X = r, Y = k r) =
k

r=0
P(X = r, Y = k r)
=
k

r=0
_
n
r
_
p
r
(1 p)
nr
_
m
k r
_
p
kr
(1 p)
m(kr)
= p
k
(1 p)
n+mk
k

r=0
_
n
r
__
m
k r
_
for k = 0, 1, . . . , n +m. Using the identity

k
r=0
_
n
r
__
m
kr
_
=
_
n+m
k
_
, it follows
that X +Y has a binomial distribution with parameters n +m and p.
(b) By P(A | B) =
P(AB)
P(B)
, it follows that
P(X = j | X +Y = k) =
P(X = j, X +Y = k)
P(X +Y = k)
.
Using the independence of X and Y ,
P(X = j, X +Y = k) = P(X = j, Y = k j)
=
_
n
j
_
p
j
(1 p)
nj
_
m
k j
_
p
kj
j(1 p)
mk+j
and so
P(X = j, X +Y = k) =
_
n
j
__
m
k j
_
p
k
(1 p)
n+mk
for 0 j k and 0 k n +m. Hence
P(X = j | X +Y = k) =
_
n
j
__
m
kj
_
p
k
(1 p)
n+mk
_
n+m
k
_
p
k
(1 p)
n+mk
=
_
n
j
__
m
kj
_
_
n+m
k
_ .
For xed k, the probabilities P(X = j | X + Y = k) for j = 0, . . . , k
constitute a hypergeometric distribution.
9E-18 For xed j and k, let A be the event that X takes on the value j and
Y takes on the value k and B the event that N takes on the value j + k.
Then, by P(AB) = P(A | B)P(B),
P(X = j, Y = k) = P(X = j, Y = k)P(N = j +k)
=
_
j +k
j
_
p
j
(1 p)
k
e


j+k
(j +k)!
= e
p
(p)
j
j!
e
(1p)
((1 p))
k
k!
.
53
Using the relations P(X = j) =

k=0
P(X = j, Y = k) and P(Y = k) =

j=0
P(X = j, Y = k), it next follows that
P(X = j) = e
p
(p)
j
j!
and P(Y = k) = e
(1p)
((1 p))
k
k!
for all j 0 and k 0. Hence X and Y are Poisson distributed with
respective expected values p and (1 p). Moreover,
P(X = j, Y = k) = P(X = j)P(Y = k) for all j, k,
showing that X and Y are independent.
9E-19 (a) By the substitution rule,
E[g(X + 1) Xg(X)]
=

k=0
g(k + 1)e

k
k!

k=0
kg(k)e

k
k!
=

k=0
g(k + 1)e

k
k!

l=0
g(l + 1)e

l
l!
= 0.
(b) Let p
j
= P(X = j) for j = 0, 1, . . .. For xed i 1, dene the indicator
function g(x) by g(k) = 1 for k = i and g(k) = 0 for k = i. Then the relation
E[g(X + 1) Xg(X)] = 0 reduces to
p
i1
ip
i
= 0.
This gives p
i
=

i
p
i1
for i 0. By repeated application of this equation, it
next follows that p
i
=

i
i!
p
0
for i 0. Using the fact that

i=0
p
i
= 1, we
get p
0
= e

. This gives
P(X = i) = e

i
i!
for i = 0, 1, . . . ,
proving the desired result.
9E-20 Let the random variable N be the outcome of the initial roll of the
die. Using the law of conditional probability with the conditioning events
B
j
= {N = j}, it follows that
P(X = k) =
6

j=1
P(X = k | N = j)P(N = j), k = 0, 1, . . . , 7.
54
Under the condition that the rst roll gives the outcome j, the probability
of a total k sixes is equal to the binomial probability
_
j
k
_
(1/6)
k
(5/6)
jk
if
1 j 5 and is equal to the binomial probability
_
6
k1
_
(1/6)
k1
(5/6)
6k+1
if j = 6, with the convention that
_
j
k
_
= 0 for k > j. This results in
P(X = 0) =
1
6

5
j=1
(5/6)
j
and
P(X = k) =
1
6
5

j=k
_
j
k
__
1
6
_
k
_
5
6
_
jk
+
1
6
_
6
k 1
__
1
6
_
k1
_
5
6
_
6k+1
, k = 1, 2, . . . , 7.
The probability P(X = k) has the numerical values 0.4984, 0.3190, 0.1293,
0.0422, 0.0096, 0.0014, and 3.57 10
6
for k = 0, 1, 2, 3, 4, 5, 6, and 7.
9E-21 Let the random variable X be your payo in any play. Noting that
the probability of drawing j gold-colored balls is
p
j
=
_
4
j
__
6
3j
_
_
10
4
_ for j = 0, 1, 2, 3,
it follows that E(X) = 1 p
2
+11 p
3
= 1
3
10
+11
1
30
=
2
3
dollars. Hence
your expected loss in any play is
2
3
=
1
3
dollars.
9E-22 To nd the two depleted batteries, you need no more than four tests
(why?). Obviously, P(X = 0) = P(X = 1) = 0. You need two tests if the
rst two batteries tested are depleted and so
P(X = 2) =
_
2
2
_
_
5
2
_ =
1
10
.
You need three tests if the rst three batteries tested are not depleted or if
a second depleted battery is found at the third test. This gives
P(X = 3) =
_
3
3
_
_
5
3
_ +
_
2
1
__
3
1
_
_
5
2
_
1
3
=
3
10
.
Similarly, you need four tests only if there is one depleted battery among the
rst three batteries tested and so
P(X = 4) =
_
2
1
__
3
2
_
_
5
3
_ =
6
10
.
55
9E-23 The problem can be translated into the urn model with 1,422 white
balls and 1,405 black balls. The desired probability is equal to the probability
the number of white balls remaining is smaller than or equal the number of
black balls remaining when 101 randomly chosen balls are removed from the
urn. Noting that 1,422 m 1,405 (101 m) only if m 59, it follows
that this probability is given by
101

m=59
_
1,422
m
__
1,405
101m
_
_
2,827
101
_ = 0.05917.
9E-24 Let E
0
be the event that you get k 1 red balls in the rst k 1 draws
and E
1
be the event of a red ball at the kth draw. The desired probability
is P(E
0
E
1
). It holds that
P(E
0
) =
_
n
k1
__
m
k1(r1)
_
_
n+m
k1
_ and P(E
1
| E
0
) =
n (r 1)
n +m(k 1)
Hence, by P(E
0
E
1
) = P(E
0
)P(E
1
| E
0
), the desired probability is given by
_
n
k1
__
m
kr
_
_
n+m
k1
_
n r + 1
n +mk + 1
for k = r, . . . , m+r.
9E-25 Assuming that the game is not stopped, let the random variable X
1
be the number of draws you need to get one of the numbers 1, . . . , 4 and
X
2
be the number of draws your friend needs to get one of the numbers
5, . . . , 10. The probability that you will be the winner is P(X
1
X
2
). Since
you and your friend draw the random numbers independently of each other
with respective success probabilities p
1
=
4
10
and p
2
=
6
10
, it follows that
P(X
1
X
2
) =

j=1
P(X
1
= j, X
2
j)
=

j=1
p
1
(1 p
1
)
j1
(1 p
2
)
j1
=
p
1
p
1
+p
2
p
1
p
2
.
Hence your probability of winning is 0.5263. It is a surprising result that
your probability of winning is more than 50%, while p
1
is smaller than p
2
.
The length of the game is min(X
1
, X
2
). It holds that
P(min(X
1
, X
2
) > l) = P(X
1
> l, X
2
> l)
= (1 p
1
)
l
(1 p
2
)
l
= (1 p)
l
, l = 0, 1, . . . ,
56
where p = p
1
+ p
2
p
1
p
2
. Using the relation P(min(X
1
, X
2
) = l) =
P(min(X
1
, X
2
) > l 1) P(min(X
1
, X
2
) > l), it next follows that
P(min(X
1
, X
2
) = l) = p(1 p)
l1
for l = 1, 2, . . . .
In other words, the length of the game is geometrically distributed with
parameter p
1
+p
2
p
1
p
2
= 0.76.
9E-26 Let the random variable X be the total demand for spare parts in the
remaining lifetime of the airplane. Using the law of conditional probabilities
and noting that the sum of independent Poisson random variables is again
Poisson distributed, it follows that
P(X = k) =
1
3
5

l=3
e
l
(l)
k
k!
for k = 0, 1, . . . .
The probability that there will be a shortage in the remaining lifetime of the
plane is

k=Q+1
P(X = k) =
1
3

k=Q+1
5

l=3
e
l
(l)
k
k!
=
1
3
5

l=3
e
l

k=Q+1
(l)
k
k!
Using the fact that

k=Q+1
(l)
k
k!
= e
l

Q
k=0
(l)
k
k!
, it follows that
P(shortage) =
1
3
5

l=3
_
1
Q

k=0
e
l
(l)
k
k!
_
.
The number of spare parts left over at the end of the remaining lifetime of
the plane is max(QX, 0). By the substitution rule,
E(number of spare parts left over) =
Q

k=0
(Qk)P(X = k)
=
1
3
5

l=3
e
l
Q

k=0
(Qk)
(l)
k
k!
.
The amount of demand that cannot be satised from the stock of Q units is
57
max(X Q, 0). By the substitution rule,
E(shortage) =

k=Q+1
(k Q)P(X = k)
=

k=0
(k Q)P(X = k)
Q

k=0
(k Q)P(X = k)
=
1
3
(3 + 4 + 5) Q+
1
3
5

l=3
e
l
Q

k=0
(Qk)
(l)
k
k!
.
9E-27 Let E be the event that no more than one of your ten trees will not
grow well and B
k
be the event that there are k trees of tree-nurseryman A
are among your ten trees. By the law of conditional probability, P(E) =

10
k=0
P(E | B
k
)P(B
k
). For any k = 0, 1, . . . , 10,
P(B
k
) =
_
50
k
__
50
10k
_
_
100
10
_ .
The conditional probability P(A | B
k
) is given by
P(A | B
k
) = (0.90)
k
(0.95)
10k
+
_
k
1
_
0.10 (0.90)
k1
0.95
10k
+
_
10 k
1
_
0.05 (0.95)
9k
0.90
k
for k = 0, 1, . . . , 10
with the convention
_
0
1
_
= 0. Substituting the expressions for P(E | B
k
) and
P(B
k
) into P(E) =

10
k=0
P(E | B
k
)P(B
k
) gives
P(E) = 0.8243.
Chapter 10
10E-1 The requirement 1 = c
_
1
0
(x +

x) dx =
7
6
c gives c =
6
7
. To nd the
density function of Y =
1
X
, we determine P(Y y). Obviously, P(Y y) =
0 for y 1. For y > 1,
P(Y y) = P
_
X
1
y
_
= 1 P
_
X
1
y
_
= 1 F
_
1
y
_
,
where F(x) is the probability distribution function of X. By dierentiation,
it follows that the density function g(y) of Y is given by
g(y) =
6
7
f
_
1
y
_

1
y
2
=
6
7
_
1
y
3
+
1
y
2

y
_
for y > 1
58
and g(y) = 0 otherwise.
10E-2 The area of the circle is Y = 2X
2
, where X is uniformly distributed
on (0, 1). The density function of X is given by f(x) = 1 for 0 < x < 1 and
f(x) = 0 otherwise. For any y with 0 y 2,
P(Y y) = P
_
X
_
y
2
_
=
_
y/2
0
dx =
_
y
2
.
Dierentiation gives that the density function of the area of the circle is given
by g(y) =
1
2

2y
for 0 y 2 and g(y) = 0 otherwise.
10E-3 The range of the random variable X is the interval [0, 1]. Let A be the
subset of points from the rectangle for which the distance to the closest side
of the rectangle is larger than x, where 0 x 1. Then A is a rectangle
whose sides have the lengths 3 2x and 2 2x and so the area of A is
(3 2x)(2 2x). It now follows that
P(X x) =
6 (3 2x)(2 2x)
6
=
5
3
x
2
3
x
2
for 0 x 1.
The probability density f(x) of X is given by f(x) =
5
3

4
3
x for 0 < x < 1
and f(x) = 0 otherwise. The expected value of X is
_
1
0
x
_
5
3

4
3
x
_
dx =
7
18
. Noting that E(X
2
) =
_
1
0
x
2
_
5
3

4
3
x
_
dx =
2
9
, it follows that (X) =
_
2/9 (7/18)
2
=
1
18

23.
10E-4 The probability distribution function of the weekly volume of waste in
thousands of gallons is given by
F(x) = 105
_
x
0
y
4
(1 y)
2
dy = 105
_
x
0
(y
4
2y
5
+y
6
) dy
= x
5
(15x
2
35x + 21) for 0 x 1.
The solution of the equation 1 F(x) = 0.05 is x = 0.8712. Hence the
capacity of the storage tank in thousands of gallons should be 0.8712.
10E-5 Let random variable X be the amount of waste (in thousands of gal-
lons) produced during a week and Y be the total costs incurred during a
week. Then the random variable Y can be represented as Y = g(X), where
the function g(x) is given by
g(x) =
_
1.25 + 0.5x for 0 < x < 0.9,
1.25 + 0.5 0.9 + 5 + 10(x 0.9) for 0.9 < x < 1
59
and g(x) = 0 otherwise. By the substitution rule, the expected value of the
weekly costs is given by
105
_
1
0
g(x)x
4
(1 x)
2
dx = 1.69747.
To nd the standard deviation of the weekly costs, we rst calculate
E(Y
2
) =
_
1
0
g
2
(x)x
4
(1 x)
2
dx = 3.62044.
Hence the standard deviation of the weekly costs is equal to

3.62044 1.69747
2
=
0.8597.
10E-6 A stockout occurs if the demand X is larger than Q and so
P(stockout) =
_

Q
f(x) dx = 1
_
Q
0
f(x) dx.
The amount of stock left over at the end of the period is max(Q X, 0).
By the substitution rule, we can compute the expected value of the random
variable Y = max(QX, 0). This gives
E(stock left over) =
_
Q
0
(Qx)f(x) dx.
The amount that cannot be satised from stock is max(X Q, 0). By the
substitution rule,
E(shortage) =
_

Q
(x Q)f(x) dx = Q+
_
Q
0
(Qx)f(x) dx,
where = E(X).
10E-7 Denoting by the point O the center of the circle, let the random variable
be the angle between the line segments PO and QO. Then is uniformly
distributed between 0 and . A little geometry shows that sin(
1
2
) =
X/2
r
and so X = 2r sin(
1
2
). Hence, by the substitution rule,
E(X) =
_

0
2r sin
_
1
2

_
1

d =
4r

_
/2
0
sin(x) dx
=
4r

cos(x)

/2
0
=
4r

1.273r.
60
Noting that sin
2
(x) + cos
2
(x) = 1 and
_
/2
0
sin
2
(x) dx =
_
/2
0
cos
2
(x) dx, it
follows that
_
/2
0
sin
2
(x) dx =

4
. Hence, by using again the substitution rule,
E(X
2
) =
_

0
4r
2
sin
2
_
1
2

_
1

d =
8r
2

_
/2
0
sin
2
(x) dx
=
8r
2



4
= 2r
2
.
This gives
(X) =
_
2r
2

16r
2

2
0.616r.
10E-8 Let the random variable X be the amount of damage. Then Y =
g(X), where the function g(x) is given by g(x) = 0 for x 450 and g(x) =
min(500, x 450) for x > 450. Using the substitution rule, it follows that
E(Y ) =
_
950
450
(x 450)
1
1,000
dx + 500
_
1,250
950
1
1,000
dx = 275.
The probability distribution function of Y is given by
P(Y = 0) = P(250 X 450) = 0.20,
P(Y y) = P(Y = 0) +
_
450+y
450
1
1,000
dx = 0.20 +
y
1,000
for y < 500,
P(Y = 500) = P(950 < X 1,250) = 0.30.
The random variable Y has a mixed distribution with a probability mass at
each of the points 0 and 500 and a density on the interval (0, 500).
10E-9 Let the N(7, 4)-distributed random variable X be the lifetime of the
item. The expected value of the payment per insurance is
aP(X 2) +
1
2
aP(2 < X 4)
= aP
_
X 7
2

2 7
2
_
+
1
2
aP
_
2 7
2
<
X 7
2

4 7
2
_
= a(2.5) +
1
2
a[(1.5) (2.5)] = a 0.03651.
Hence the value 1370 should be taken for a.
10E-10 The area of the triangle is given by V =
1
2
tg(). The density f()
of is given by f() =
4

for 0 < <



4
and f() = 0 otherwise. By the
substitution rule, the rst two moments of V are given by
E(V ) =
1
2
_
/4
0
tg()
4

d =
1

ln(2)
61
and
E(V
2
) =
1
4
_
/4
0
tg
2
()
4

d =
1

_
1

4
_
.
10E-11 The quadratic distance is distributed as X
2
+Y
2
, where X and Y are
independent random variables each having the standard normal density. The
random variable V = X
2
+Y
2
has the chi-square density with two degrees of
freedom. This density is given by f(v) =
1
2
e

1
2
v
for v > 0. Denoting by F(v)
the probability distribution function of the
2
2
distributed random variable
V , it follows that the probability distribution function of the distance from
the center of the target to the point of impact is given by
P(

X
2
+Y
2
r) = P(V r
2
) = F(r
2
) for r > 0.
Hence the probability density of the distance from the center of the target to
the point of impact is 2rf(r
2
) = re

1
2
r
2
for r > 0. The expected value of the
distance is
_

0
r
2
e

1
2
r
2
dr =
1
2

2, using the fact that


1

2
_

r
2
e

1
2
r
2
dr = 1
for the standard normal density. The mode of the distance follows by solving
the equation
_
x
0
re

1
2
r
2
dr = 0.5. This gives the value
_
2 ln(2) for the mode.
10E-12 The radius of the circle is given by the random variable R =

X
2
+Y
2
and the area of the circle is (X
2
+ Y
2
). Using the fact that X
2
+ Y
2
has
the
2
2
-density f(x) =
1
2
e

1
2
x
for x > 0, it follows that
P
_
(X
2
+Y
2
) v
_
= P
_
X
2
+Y
2

_
=
_
v/
0
1
2
e

1
2
w
dw, v > 0.
Hence the density function of the area of the circle is
1
2
e

1
2
v/
for v > 0.
This is the exponential density with parameter 2. The expected value of
the area of the circle is
_

0
v
1
2
e

1
2
v/
dv = 2.
10E-13 Let the random variable X denote your service time, For xed t > 0,
let A be the event that your service time is no more than t and B
i
be the event
that you are routed to server i for i = 1, 2. Then, by the law of conditional
probability, P(A) = P(A | B
1
)P(B
1
) +P(A | B
2
)P(B
2
). Using the fact that
P(V t) = 1e
t
for an exponentially distributed random variable V with
parameter , it now follows that
P(X t) = p
1
(1 e

1
t
) +p
2
(1 e

2
t
).
Dierentiation gives that the density function f(t) of X is given by
f(t) = p
1

1
e

1
t
+p
2

2
e

2
t
for t > 0.
62
This is the so-called hyperexponential density.
10E-14 (a) The probability distribution function of Y is given by
P(Y y) = P(

y X

y)
=
1

2
_

y

1
2
x
2
dx
1

2
_

1
2
x
2
dx.
Dierentiation of P(Y y) gives that the probability density function of Y
is given by
g(y) =
1

2
e

1
2
y
1
2

y
+
1

2
e

1
2
y
1
2

y
=
1

2
y
1
2
e

1
2
y
for y 0
and g(y) = 0 otherwise. This shows that Y has a gamma density with shape
parameter
1
2
and scale parameter
1
2
.
(b) Noting that P(Y y) = P(|X| y
2
) for y 0, it follows that
P(Y y) = P(y
2
X y
2
) = (y
2
) (y
2
) for y 0.
Put for abbreviation (x) =
1

2
e

1
2
x
2
. Dierentiation of the distribution
function P(Y y) gives that the density function of Y is equal to
2y(y
2
) + 2y(y
2
) =
4y

2
e

1
2
y
4
for y > 0
and is equal to zero otherwise.
10E-15 The probability distribution function of Y is given by
P(Y y) = P(2X 1 y) = P
_
X
1
2
(y + 1)
_
=
_ 1
2
(y+1)
0
8

_
x(1 x) dx, 1 y 1.
The derivative of P(Y y) is (8/)
_
1
2
(y + 1)(1
1
2
(y + 1))
1
2
=
2

_
1 y
2
for 1 < y < 1. Hence the probability density function of Y is given by
g(y) =
2

_
1 y
2
) for 1 < y < 1 and g(y) = 0 otherwise. This is the
so-called semicircle density.
10E-16 The probability density function g(y) of the random variable Y is
given by f(a(y))|a

(y)|, where the inverse function a(y) =


1
y
. This gives
g(y) =
1
(1 + (1/y)
2
)

1
y
2

=
1
(1 +y
2
)
for < y < .
63
In other words, the random variable
1
X
has the same Cauchy distribution as
X.
10E-17 Let us rst dene the conditional density function of X given that
X > a. To do so, we determine P(X x | X a) By P(A | B) =
P(AB)/P(B), we have for any x < a that
P(X x | X a) =
P(X x)
P(X a)
=
_
x

f(y) dy
_
a

f(y) dy
.
Obviously, P(X x | X a) = 1 for x a. This explains why conditional
probability density f
a
(x) of X given that X a is dened by
f
a
(x) =
f(x)
_
a

f(y) dy
for x < a
and f
a
(x) = 0 otherwise. The conditional expected value E(X | X a) is
dened by
E(X | X a) =
_
a

xf
a
(x) dx.
For the case that x has the exponential density with parameter ,
E(X | X a) =
_
a

xe
x
dx
_
a

e
x
dx
=
1 e
a
ae
a
1 e
a
.
10E-18 The probability that the time between the passings of two consecutive
cars is more than c seconds is given by
p =
_

c
e
t
dt = e
c
.
By the lack of memory of the exponential distribution, the probability p =
e
c
gives also the probability that no car comes around the corner during
the c seconds measured from the moment you arrive at the road. Denoting
by the random variable N the number of passing cars before you can cross
the road, it now follows that N has the shifted geometric distribution
P(X = k) = (1 p)
k
p for k = 0, 1, . . . .
10E-19 By the lack of memory of the exponential distribution, the remaining
washing time of the car being washed in the station has the same exponential
density as a newly started washing time. Hence the probability that the car
in the washing station will need no more than ve other minutes is equal to
_
5
0
1
15
e

1
15
t
dt = 1 e
5/15
= 0.2835.
64
The probability that you have to wait more than 20 minutes before your car
can be washed is equal to the probability of 0 service completions or 1 service
completion in the coming 20 minutes. The latter probability is given by the
Poisson probability (see Rule 10.2 in the book)
e
20/15
+
20
15
e
20/15
= 0.6151.
10E-20 The probability that the closest integer to the random observation is
odd is equal to

k=0
P(2k +
1
2
< X < 2k + 1 +
1
2
) =

k=0
_
2k+1+
1
2
2k+
1
2
e
x
dx
=

k=0
_
e
(2k+
1
2
)
e
(2k+1+
1
2
)

= e

1
2
_
1 e
1
1 e
2
_
=
e

1
2
1 +e
1
.
The conditional probability that the closest integer to the random observation
is odd given that it is larger than the even integer r is equal to

k=0
P(2k +
1
2
< X < 2k + 1 +
1
2

X > r)
=
1
P(X > r)

k=0
P(2k +
1
2
< X < 2k + 1 +
1
2
, X > r)
=
1
e
r

k=r/2
_
2k+1+
1
2
2k+
1
2
e
x
dx =
1
e
r

k=r/2
_
e
(2k+
1
2
)
e
(2k+1+
1
2
)

Since

k=r/2
_
e
(2k+
1
2
)
e
(2k+1+
1
2
)

= e
r

l=0
_
e
(2l+
1
2
)
e
(2l+1+
1
2
)

, the
conditional probability that the closest integer to the random observation is
odd given that it is larger than r is equal to

l=0
_
e
(2l+
1
2
)
e
(2l+1+
1
2
)

=
e

1
2
1 +e
1
.
The unconditional probability is the same as the unconditional probability
that the closest integer to the random observation from the exponential den-
sity is odd. This result can also be explained from the memoryless property
of the exponential distribution.
10E-21 Your probability of winning is the probability of having exactly one
signal in the time interval (s, T). By the memoryless property of the Poisson
65
process, this probability is equal to e
(Ts)
(T s). Putting the derivative
of this expression equal to zero, it follows that the optimal value of s is given
by T
1

. The maximal win probability is e


1
( 0.3679), irrespective of the
values of and T.
10E-22 Let p be the probability of no car passing through the village during
the next half hour. Then, by the memoryless property of the Poisson process,
the probability of no car passing through the village during one hour is p
2
.
Solving p
2
= 1 0.64 gives p = 0.6. Hence the desired probability is 0.4.
10E-23 Let the random variables X
1
and X
2
be the two measurement errors.
The desired probability is P
_
|
X
1
+X
2
2
| 0.005l
_
. Since the random variables
X
1
and X
2
are independent, the random variable
1
2
(X
1
+ X
2
) is normally
distributed with mean 0 and standard deviation
1
2

0.006
2
l
2
+ 0.004
2
l
2
=
l
2,000

52. Letting Z be standard normally distributed, it now follows that


P
_

X
1
+X
2
2

0.005l
_
= P
_

X
1
+X
2
0
2(l/2,000)

52


1,000 0.005

52
_
= P
_
|Z|
10

52
_
= P
_

10

52
Z
10

52
_
.
Hence the desired probability is given by
P
_

X
1
+X
2
2

0.005l
_
=
_
10

52
_

10

52
_
= 2
_
10

52
_
1 = 0.8345.
10E-24 The desired probability is P(|X
1
X
2
| a). Since the random
variables X
1
and X
2
are independent, the random variable X
1
X
2
is normal
distributed with expected value =
1

2
and standard deviation =
_

2
1
+
2
2
. It now follows that
P(|X
1
X
2
| a) = P(a X
1
X
2
a)
= P
_
a


X
1
X
2

_
.
Hence, denoting by (x) the standard normal distribution function,
P(|X
1
X
2
| a) =
_
a (
1

2
)
_

2
1
+
2
2
_

_
a (
1

2
)
_

2
1
+
2
2
_
.
10E-25 The probability mass function of the number of copies of the appliance
to be used when an innite supply would be available is a Poisson distribution
66
with expected value of
150
2
= 75. Suppose that Q copies of the appliance are
stored in the space ship. Let the exponentially distributed random variable
X
i
be the lifetime (in days) of the ith copy used. Then the probability of a
shortage during the space mission is P(X
1
+ + X
Q
150). The random
variables X
1
, . . . , X
Q
are independent and have an expected value of
1

days
and a standard deviation of
1

days, where =
1
2
. By the central limit
theorem,
P(X
1
+ +X
Q
150) = P
_
X
1
+ +X
Q
2Q
2

Q

150 2Q
2

Q
_

_
150 2Q
2

Q
_
.
The 0.001 percentile of the standard normal distribution is -3.0902. Solv-
ing the equation
1502Q
2

Q
= 3.0902 gives Q = 106.96 and so the nor-
mal approximation suggests to store 107 units. The exact value of the re-
quired stock follows by nding the smallest value of Q for which the Pois-
son probability

k>Q
e
75 75
k
k!
is smaller than or equal to 10
3
. This gives
Q = 103.
10E-26 Let X
i
be the amount (in dollars) the casino owner loses on the ith
bet. Then X
1
, . . . , X
2,500
are independent random variables with P(X
i
=
10) =
18
37
and P(X
i
= 5) =
19
37
for 1 i 2,500. The amount (in dollars)
lost by the casino owner is X
1
+ +X
2,500
. For any i,
E(X
i
) =
85
37
= 2.29730 and (X
i
) =

2275
37

_
85
37
_
2
= 7.49726.
By the central limit theorem, X
1
+ + X
2,500
is approximately N(,
2
)
distributed with = 2.2973022,500 = 5,743.250 and = 7.49726

2,500 =
374.863. Hence the probability that the casino owner will lose more than
6,500 dollars is approximately equal to 1
_
6,5005,743.250
374.863
_
= 0.0217.
10E-27 Let the random variable V
n
be the bankroll (in dollars) of the gambler
after the nth bet. Then V
n
= (1 )V
n1
+V
n1
R
n
for 1 n 100, where
= 0.05, V
0
= 1,000 and R
1
, . . . , R
100
are independent random variables
with P(R
i
=
1
4
) =
19
37
and P(R
i
= 2) =
18
37
for all i. Iterating this equality
gives
V
n
= (1 +R
1
) (1 +R
n
)V
0
for n = 1, 2, . . . , 100.
Taking logarithms, we get
ln(V
n
/V
0
) = ln(1 +R
1
) + + ln(1 +R
n
) for all n.
67
Let X
i
= ln(1+R
i
). The random variables X
1
, . . . , X
100
are independent
with
E(X
i
) =
19
37
ln(0.9625) +
18
37
ln(1.05) = 0.0041086
and
(X
i
) =
_
19
37
ln
2
(0.9625) +
18
37
ln
2
(1.05) 0.0041086
2
= 0.0434898.
By the central limit theorem, ln(V
100
/V
0
) is approximately N(,
2
) dis-
tributed, where
= 0.0434898 100 = 0.41086 and = 0.0434898

100 = 0.434898.
To approximate the probability that the gambler takes home more than d
dollars for d = 0, 500, 1,000 and 2,500, we consider P(V
n
> aV
0
) for a = 1,
1.5, 2 and 3.5. We have
P(V
n
> aV
0
) = P(ln(V
n
/V
0
) > ln(a))
= P
_
ln(V
n
/V
0
)

>
ln(a)

_
and so
P(V
n
> aV
0
) 1
_
ln(a)

_
.
The probability that the gambler takes home more than d dollars has the
values 0.8276, 0.5494, 0.2581, and 0.0264 for d = 0, 500, 1,000 and 2,500.
Note: the random variable V
100
being the gamblers bankroll after 100 bets
is approximately lognormally distributed with an expected value of e
+
1
2

1,000 = 1,000.51 dollars and a standard deviation of e


+
1
2

2
1
1,000 = 43.73 dollars.
10E-28 Taking the derivative of the failure rate function r(x), it is a matter of
simple algebra to verify that the failure rate function is increasing on (0, x

)
and decreasing on (x

, ), where x

is the unique solution to the equation

2
2
e

1
x
+
2
1
e

2
x
= (
2

1
)
2
.
The failure rate function determines uniquely the probability distribution
function F(x) of the lifetime of the vacuum tube. Using the basic relation
r(x) =
f(x)
1F(x
, it follows that
1 F(x) = e

1
x
+e

2
x
e
(
1
+
2
)x
for x 0.
68
Chapter 11
11E-1 (a) The joint probability mass function is given by P(X = 0, Y = 0) =
1
8
, P(X = 0, Y = 1) =
1
8
, P(X = 1, Y = 0) =
1
8
, P(X = 1, Y = 1) =
1
4
,
P(X = 1, Y = 2) =
1
8
, P(X = 2, Y = 1) =
1
8
, and P(X = 2, Y = 2) =
1
8
.
Using the relation E(XY ) =

2
x=0

2
y=0
)P(X = x, Y = y), it follows that
E(XY ) = 1
1
4
+ 2
1
8
+ 2
1
8
+ 4
1
8
= 1.25.
(b) The joint probability mass function is given by
P(X = x, Y = y) =
_
5
x
__
1
2
_
5
_
x
y
__
1
2
_
x
for 0 x 5, 0 y x.
Using the relation E(X + Y ) =

5
x=0

x
y=0
(x + y)P(X = x, Y = y), it
follows that
E(X +Y ) =
5

x=0
(x + 0.5x)
_
5
x
__
1
2
_
5
= 2.5 + 1.25 = 3.75.
11E-2 The joint probability mass function is given by
P(X = 1, Y = k) =
_
k 1
3
_
(0.55)
4
(0.45)
k4
for k = 4, . . . , 7.
and
P(X = 0, Y = k) =
_
k 1
3
_
(0.45)
4
(0.55)
k4
for k = 4, . . . , 7.
11E-3 To nd the joint probability mass function of X and Y , we rst cal-
culate
P(X = x, Y = y, N = n) =
1
6
_
n
x
__
1
2
_
n
_
n
y
__
1
2
_
n
for 1 n 6 and 0 x, y n. It now follows that the joint probability
mass function of X and Y is given by
P(X = x, Y = y) =
1
6
6

n=1
_
n
x
__
n
y
__
1
2
_
2n
, 0 x, y 6,
with the convention
_
n
k
_
= 0 for k > n. To calculate the numerical value of
P(X = Y , note that
P(X = Y ) =
1
6
6

k=0
6

n=1
_
n
k
_
2
_
1
2
_
2n
=
1
6
6

n=1
_
1
2
_
2n n

k=0
_
n
k
_
2
.
69
The relation

n
k=0
_
n
k
_
2
=
_
2n
n
_
holds , as follows from
_
2n
n
_
=

n
k=0
_
n
k
__
n
nk
_
.
We can now conclude that
P(X = Y ) =
1
6
6

n=1
_
2n
n
_
2
_
1
2
_
2n
= 0.3221.
11E-4 The probability P(X = i, Y = i) is equal the probability that two or
more dice yield the highest score i and the other dice score less than i. Hence
P(X = i, Y = i) =
d

r=2
_
d
r
__
1
6
_
r
_
i 1
6
_
dr
=
d

r=0
_
d
r
__
1
6
_
r
_
i 1
6
_
dr

_
i 1
6
_
d
d
_
i 1
6
_
d1
=
i
d
(i 1)
d
d(i 1)
d1
6
d
.
The probability P(X = i, Y = j) for i > j is equal the probability that
exactly one die yields the highest score i, at least one die yields the second-
highest score j and the other dice score less than j. Hence, for i > j,
P(X = i, Y = j) =
_
d
1
_
1
6
d1

r=1
_
d 1
r
__
1
6
_
r
_
j 1
6
_
d1r
=
_
d
1
_
1
6
_
d1

r=0
_
d 1
r
__
1
6
_
r
_
j 1
6
_
d1r

_
j 1
6
_
d1
_
=
d[j
d1
(j 1)
d1
]
6
d
.
11E-5 (a) The marginal distribution of X is given by
P(X = x) =

y=x
e
2
x!(y x)!
=
e
2
x!

k=0
1
k!
=
e
1
x!
for x = 0, 1, . . .
and the marginal distribution of Y is given by
P(Y = y) =
y

x=0
e
2
x!(y x)!
=
e
2
y!
y

x=0
y!
x!(y x)!
=
e
2
2
y
y!
for y = 0, 1, . . . .
(b) Let V = Y X. The joint probability mass function of X and Y X is
P(X = x, Y X = z) = P(X = x, Y = z +x) =
e
2
x!z!
for x, z = 0, 1, . . . .
70
Since P(Y X = z) =

x=0
e
2
x!z!
=
e
1
z!
for z = 0, 1, . . ., it follows that
P(X = x, Y X = z) = P(X = x)P(Y X = z) for all x, z
and so X and Y X are independent.
(c) E(XY ) can be computed from E(XY ) =

x=0

y=x
xy
e
2
x!(yx)!
, but it is
easier to compute E(XY ) from E(XY ) = EX(Y X +X)] = E(X)E(Y
X)+E(X
2
), using the independence of X and Y . Since X and Y X are both
Poisson distributed with expected value 1, it follows that E(XY ) = 1+2 = 3.
The expected value and the variance of the Poisson distributed variable X
are given by E(X) = 1 and
2
(X) = 1. The expected value and the variance
of the random variable Y are easiest computes as E(Y ) = E(Y X +X) =
E(Y X) +E(X) = 1+1 = 2 and
2
(Y ) =
2
(Y X) +
2
(X) = 1+1 = 2.
This gives
(X, Y ) =
3 1 2
1

2
=
1

2
.
11E-6 The joint probability mass function of X and Y is given by
P(X = x, Y = y) =
_
6
x
__
6x
y
_
4
6xy
6
6
for x, y 0, x +y 6.
The easiest way to compute (X, Y ) is to use indicator variables. Let I
j
= 1
if the jth roll gives the outcome 1 and I
j
= 0 otherwise. Also, let L
j
= 1 if
the jth roll gives the outcome 6 and L
j
= 0 otherwise. Then X =

6
j=1
I
j
and Y =

6
j=1
L
j
. For each j, E(I
j
) = E(L
j
) =
1
6
and
2
(I
j
) =
2
(L
j
) =
5
36
.
This gives E(X) = E(Y ) = 1 and
2
(X) =
2
(Y ) =
5
6
, where the latter
expression uses the fact that both I
1
, . . . , I
6
and L
1
, . . . L
6
are independent
random variables. To nd E(XY ), note that I
j
L
j
= 0 for all j and I
j
is
independent of L
k
for j = k. This gives
E(XY ) = E
_
6

j=1
I
j
6

k=1
L
k
_
=
6

j=1

k=j
E(I
j
)E(L
k
) =
30
36
=
5
6
.
We can now conclude that
(X, Y ) =
5/6 1 1
_
5/6
_
5/6
= 0.2.
11E-7 The constant c follows from the requirement c
_
1
0
dx
_
x
0
xy dy = 1. This
gives c
_
1
0
1
2
x
3
dx = 1 and so c = 8. The marginal density of X is given by
f
X
(x) =
_
x
0
8xy dy = 4x
3
for 0 < x < 1
71
and f
X
(x) = 0 otherwise. The marginal density of Y is
f
Y
(y) =
_
1
y
8xy dx = 4y(1 y
2
) for 0 < y < 1
and f
Y
(y) = 0 otherwise.
11E-8 The radius of the circle is

X
2
+Y
2
and so its circumference is given
by V = 2

X
2
+Y
2
. To nd the desired probability, use the relation
P((X, Y ) C) =
__
C
f(x, y) dxdy. This gives
P(V 2) = P(X
2
+Y
2
1) =
_
1
0
dx
_

1x
2
0
(x +y)dy
=
_
1
0
_
x

1 x
2
+
1
2
(1 x
2
)
_
dx =
1
3
+
1
3
=
2
3
.
11E-9 Using the basic formula P((X, Y ) C) =
__
C
f(x, y) dxdy, it follows
that
P(X < Y ) =
1
10
_
10
5
dx
_

x
e

1
2
(y+3x)
dy
=
2
10
_
10
5
e

1
2
(3x)
e

1
2
x
= e

3
2
= 0.2231.
11E-10 Let Z = XY . Using the basic formula P((X, Y ) C) =
__
C
f(x, y) dxdy,
it follows that
P(Z z) =
_
z
0
dx
_
z/x
0
xe
x(y+1)
dy =
_
z
0
e
x
(1 e
xz/x
) dx =
= (1 e
z
)(1 e
z
) for z 0.
Hence the density function of Z = XY is given by f(z) = 2e
z
(1 e
z
) for
z > 0 and f(z) = 0 otherwise.
11E-11 Let Z = XY . Using the basic formula P((X, Y ) C) =
__
C
f(x, y) dxdy,
it follows that
P(Z z) =
1
2
_
z
0
dx
_
zx
0
(x +y)e
(x+y)
dy =
1
2
_
z
0
dx
_
z
x
ue
u
du
=
1
2
_
z
0
(ze
z
+xe
x
+e
x
e
z
) dx = 1 e
z
(1 +z +
1
2
z
2
)
for z 0. Hence the density function of Z = X + Y is f(z) =
1
2
z
2
e
z
for z > 0 and f(z) = 0 otherwise. This is the Erlang density with shape
parameter 3 and scale parameter 1.
72
11E-12 (a) Let T be the time until neither of two components is still working.
Then T = max(X, Y ). To evaluate P(T t) = P(X t, Y t), we
distinguish between 0 t 1 and 1 t 2. For 0 t 1,
P(X t, Y t) =
1
4
_
t
0
dx
_
t
0
(2y + 2 x) dy
=
1
4
_
t
0
(t
2
+ 2t xt) dx = 0.25t
3
+ 0.375t
2
.
For 1 t 2,
P(X t, Y t) =
1
4
_
t
0
dx
_
1
0
(2y + 2 x) dy
=
1
4
_
t
0
(3 x) dx = 0.75t 0.125t
2
.
The density function of T is given by 0.75(t
2
+t) for 0 < t < 1 and 0.750.25t
for 1 t < 2.
(b) The amount of time that the lifetime X survives the lifetime Y is given
by V = max(X, Y ) Y . The random variable V is a mixed random variable
with a probability mass at v = 0 and a density on (0, 2). The probability
P(V = 0) is given by P(Y X) and so
P(V = 0) =
1
4
_
1
0
dx
_
1
x
(2y + 2 x) dy
=
1
4
_
1
0
(3 3x) dx =
3
8
.
To nd the density of V on the interval (0, 2), we distinguish between the
cases 0 < v < 1 and 1 v < 2. It is easiest to evaluate P(V > v). For
0 < v < 1,
P(V > v) =
1
4
_
1
0
dy
_
1
y+v
(2y + 2 x) dx
=
1
4
_
1
0
(2y 1.5y
2
yv + 2 2v + 0.5v
2
) dy =
5
8

5
8
v +
1
8
v
2
.
For 1 v < 2,
P(V > v) =
1
4
_
2v
0
dy
_
1
y+v
(2y + 2 x) dx
=
1
4
_
2v
0
(2y 1.5y
2
yv + 2 2v + 0.5v
2
) dy
=
1
4
_
0.5v
3
+ 3v
2
6v + 4
_
.
73
Hence the density function of V on (0, 2) is given by
5
8

2
8
v for 0 < v < 1
and by
3
8
v
2
1.5v + 1.5 for 1 v < 2.
11E-13 (a) Using the basic formula P((X, Y ) C) =
__
C
f(x, y) dxdy, it
follows that
P(B
2
4A) =
_
1
0
_
1
0
(a, b)f(a, b) da db,
where the function (a, b) = 1 for b
2
4a and (a, b) = 0 otherwise. This
leads to the desired probability
P(B
2
4A) =
_
1
0
db
_
b
2
/4
0
(a +b) da =
_
1
0
_
b
4
32
+
b
3
4
_
db = 0.06875.
(b) Letting the function (a, b, c) = 1 for b
2
4ac and (a, b, c) = 0 other-
wise, it follows that
P(B
2
4AC) =
_
1
0
_
1
0
_
1
0
(a, b, c)f(a, b, c) da db dc.
Any order of integration can be used to evaluate this three-dimensional in-
tegral. This leads to
P(B
2
4AC) =
_
1
0
db
_
b
2
/4
0
db
_
1
0
(a, b, c)f(a, b, c) dc
+
_
1
0
db
_
1
b
2
/4
db
_
1
0
(a, b, c)f(a, b, c) dc.
Next it is readily seen that
P(B
2
4AC) =
2
3
_
1
0
db
_
b
2
/4
0
da
_
1
0
(a +b +c) dc
+
2
3
_
1
0
db
_
1
b
2
/4
da
_
b
2
/(4a)
0
(a +b +c) dc.
We nd
_
1
0
db
_
b
2
/4
0
da
_
1
0
(a +b +c) dc =
_
1
0
_
b
4
32
+
b
3
4
+
b
2
8
_
db = 0.110417
and
_
1
0
db
_
1
b
2
/4
da
_
b
2
/(4a)
0
(a +b +c) dc
=
_
1
0
_
_
1
b
2
4
__
b
2
4

b
4
32
_

b
3
4
ln
_
b
2
4
_
_
db.
74
Using partial integration the integral
_
1
0
b
3
4
ln
_
b
2
4
_
db can be evaluated as
1
16
_
1
0
ln
_
b
2
4
_
db
4
=
1
16
_
b
4
ln
_
b
2
4
_

1
0

_
1
0
b
4
4
b
2
b
2
db
_
=
1
16
ln
_
1
4
_

1
24
.
This gives
_
1
0
db
_
1
b
2
/4
da
_
b
2
/(4a)
0
(a +b +c) dc = 0.183593.
Hence the desired probability that Ax
2
+ Bx + C = 0 has two real roots is
given by P(B
2
4AC) =
2
3
(0.110417 + 0.183593) = 0.19601.
11E-14 The probability of a system failure between two inspections is
P(X T, Y T) =
_
T
0
_
T
0
e
(x+y)
dxdy
=
_
T
0
e
x
dx
_
T
0
e
y
dy =
_
1 e
T
_
2
.
Let the random variable D denote the amount of time the system is down
between two inspections. Then D = T max(X, Y ) if X, Y T and D = 0
otherwise. This gives
E(D) =
_
T
0
_
T
0
_
T max(x, y)
_
e
(x+y)
dxdy
= 2
_
T
0
e
x
dx
_
x
0
(T x)e
y
dy.
This leads after some algebra to
E(D) =
_
T
0
(T x)e
x
(1 e
x
) dx = T 1.5 e
T
+ 0.5e
2T
.
11-E15 The expected value of the time until the electronic device goes down
is given by
E(X +Y ) =
_

1
_

1
(x +y)
24
(x +y)
4
dxdy
=
_

1
dx
_

1
24
(x +y)
3
dy =
_

1
12
(x + 1)
2
dx = 6.
75
To nd the density function of X + Y , we calculate P(X + Y > t) and
distinguish between 0 t 2 and t > 2. Obviously, P(X + Y > t) = 1 for
0 t 2. For the case of t > 2,
P(X +Y > t) =
_
t1
1
dx
_

tx
24
(x +y)
4
dy +
_

t1
dx
_

1
24
(x +y)
4
dy
=
_
t1
1
8
t
3
dx +
_

t1
8
(x + 1)
3
dx =
8(t 2)
t
3
+
4
t
2
.
By dierentiation, the density function g(t) of X + Y is g(t) =
24(t2)
t
4
for
t > 2 and g(t) = 0 otherwise.
11E-16 By the independence of the chosen random numbers, the joint density
function of X
1
, X
2
and X
3
is 1 1 1 = 1 for 0 < x
1
, x
2
, x
3
< 1 and 0
otherwise. Let C = {(x
1
, x
2
, x
3
) : 0 < x
1
, x
2
, x
3
< 1, 0 < x
2
+ x + 3 < x
1
}.
Then
P(X
1
> X
2
+X
3
) =
_ _ _
C
dx
1
dx
2
dx
3
=
_
1
0
dx
1
_
x
1
0
dx
2
_
x
1
x
2
0
dx
3
.
This gives
P(X
1
> X
2
+X
3
) =
_
1
0
dx
1
_
x
1
0
(x
1
x
2
) dx
2
=
_
1
0
1
2
x
2
1
dx
1
=
1
2

1
3
=
1
6
.
Since the events {X
1
> X
2
+ X
3
}, {X
2
> X
1
+ X
3
} and {X
3
> X
1
+ X
2
}
are mutually exclusive, the probability that the largest of the three random
numbers is greater than the sum of the other two is 3
1
6
=
1
2
.
Remark. More generally, let X
1
, X
2
, . . . , X
n
be independent random numbers
chosen from (0, 1), then P(X
1
> X
2
+ +X
n
) =
1
n!
for any n 2.
11E-17 The joint density of X
1
and X
2
is given by f(x
1
, x
2
) = f
X
1
(x
1
)f
X
2
(x
2
) =
1 for 0 < x
1
, x
2
< 1 and f(x
1
, x
2
) = 0 otherwise. Hence
P(X
1
+X
2
1) =
_
1
0
dx
1
_
1x
1
0
dx
2
=
1
2
.
In the same way,
P(X
1
+X
2
+X
3
1) =
_
1
0
dx
1
_
1x
1
0
dx
2
_
1x
1
x
2
0
dx
3
=
1
6
.
76
Continuing in this way,
P(X
1
+X
2
+ +X
n
1) =
1
n!
.
11E-18 To nd the joint density of V and W, we apply the transforma-
tion formula. The inverse functions x = a(v, w) and y = b(v, w) are given
a(v, w) = vw/(1+w) and b(v, w) = v/(1+w). The Jacobian J(v, w) is equal
to v/(1 +w)
2
and so the joint density of V and W is given by
f
V,W
(v, w) = 1 1 |J(v, w)| =
v
(1 +w)
2
for 0 < v < 2 and w > 0
and f
V,W
(v, w) = 0 otherwise. The marginal density of V is
f
V
(v) =
_

0
v
(1 +w)
2
dw =
1
2
v for 0 < v < 2
and f
V
(v) = 0 otherwise. The marginal density of W is given by
f
W
(w) =
_
2
0
v
(1 +w)
2
dv =
2
(1 +w)
2
for w > 0
and f
W
(w) = 0 otherwise. Since f
V,W
(v, w) = f
V
(v)f
W
(w) for all v, w, the
random variables V and W are independent.
11E-19 It is not possible to obtain the joint density of V and W by applying
directly the two-dimensional transformation rule. To apply this rule, we
write V and W as V = X +Y and W = XY , where X = Z
2
1
and Y = Z
2
2
.
The probability densities of X and Y are easily found. Note that
P(X x) = P(

x Z
1

x) = (

x) (

x) for x 0.
Denoting by (x) =
1

2
e

1
2
x
2
the standard normal density. it follows that
the density f
X
(x) of X is
f
X
(x) = (

x)
1
2
x

1
2
+(

x)
1
2
x

1
2
=
1

2
x

1
2
e

1
2
x
for x > 0.
This is a gamma density with shape parameter
1
2
and scale parameter
1
2
.
The density function f
Y
(y) of Y is the same gamma density. Since Z
1
and
Z
2
are independent, X and Y are also independent. Hence the joint density
function f
X,Y
(x, y) of X and Y is given by
f
X,Y
(x, y) = f
X
(x)f
Y
(y) =
1
2
(xy)

1
2
e

1
2
(x+y)
for x, y > 0.
77
The inverse functions x = a(v, w) and y = b(v, w) are a(v, w) =
1
2
(v + w)
and b(v, w) =
1
2
(v w). The Jacobian J(v, w) is equal to
1
2
. Next it follows
that the joint density of V and W is given by
f
V,W
(v, w) =
1
4
2
(v
2
w
2
)

1
2
e

1
2
v
for v > 0 and < w < .
Since f
V,W
(v, w) is not the product of a function of only v and a function of
only w, the random variables V and W are not independent.
11E-20 The inverse functions x = a(v, w) and y = b(v, w) are given a(v, w) =
vw and b(v, w) = v(1 w). The Jacobian J(v, w) is equal to v and so the
joint density of V and W is given by
f
V,W
(v, w) = e
vw
e
v(1w)
| v| =
2
ve
v
for v > 0 and 0 < w < 1
and f
V,W
(v, w) = 0 otherwise. The marginal density of V is
f
V
(v) =
_
1
0

2
ve
v
dw =
2
ve
v
for v > 0
and f
V
(v) = 0 otherwise. The marginal density of W is given by
f
W
(w) =
_

0

2
ve
v
dv = 1 for 0 < w < 1
and f
W
(w) = 0 otherwise. In other words, V has the Erlang-2 density
with scale parameter and W is uniformly distributed on (0, 1). Since
f
V,W
(v, w) = f
V
(v)f
W
(w) for all v, w, the random variables V and W are
independent.
11E-21 (a) To nd c, use the fact that
1
a

2
_

0
e

1
2
y
2
/a
2
dy = 1 for any a > 0.
This gives
1 = c
_

0
dx
_

0
xe

1
2
x(1+y
2
)
dy =
_

0
cxe

1
2
x
dx
_

0
e

1
2
y
2
/(1/

x)
2
dy
=
_

0
cxe

1
2
x
(1/

x)

2
1
2
dx = c
1
2

2 (1.5)/0.5
1.5
= .
This gives
c =
1

.
The marginal densities of X and Y are given by
f
X
(x) =
1

_

0
xe

1
2
x(1+y
2
)
dy =
1

2
x
1
2
e

1
2
x
for x > 0
78
and
f
Y
(y) =
1

_

0
xe

1
2
x(1+y
2
)
dx =
4/
(1 +y
2
)
2
for y > 0.
(b) Let V = Y

X and W = X. To nd the joint density of V and W,


we apply the transformation formula. The inverse functions x = a(v, w) and
y = b(v, w) are given a(v, w) = w and b(v, w) = v/

w. The Jacobian J(v, w)


is equal to 1/

w and so the joint density of V and W is given by


f
V,W
(v, w) =
1

we
w(1+v
2
/w)
1

w
=
1

we

1
2
w
e

1
2
v
2
for v, w > 0
and f
V,W
(v, w) = 0 otherwise. The densities f
V
(v) =
_

0
f
V,W
(v, w)dw and
f
W
(w) =
_

0
f
V,W
(v, w)dv are given by
f
V
(v) =
_
2

1
2
v
2
for v > 0, f
W
(w) =
1

2
w
1
2
e

1
2
w
for w > 0.
The random variable V is distributed as |Z| with Z having the standard
normal distribution and W has a gamma distribution with shape parameter
3
2
and shape parameter
1
2
. Since f
V,W
(v, w) = f
V
(v)f
W
(w) for all v, w, the
random variables V and W are independent.
11E-22 The marginal density of X is given by
f
X
(x) = 6
_
x
0
(x y) dy = 3x
2
for 0 < x < 1
and f
X
(x) = 0 otherwise. The marginal density of Y is given by
f
Y
(y) = 6
_
1
y
(x y) dx = 3y
2
6y + 3 for 0 < y < 1
and f
Y
(y) = 0 otherwise. Next we calculate
E(X) =
3
4
, (X) =
1
4

3, E(Y ) =
1
4
, (Y ) =
_
3/80
and
E(XY ) = 6
_
1
0
dx
_
x
0
xy(x y) dy =
_
1
0
x
4
dx =
1
5
.
This leads to
(X, Y ) =
1/5 (3/4) (1/4)
(1/4)

3
_
3/80
= 0.1491.
79
11E-23 The linear least square estimate of D
1
given that D
1
D
2
= d is
E(D
1
) +(D
1
D
2
, D
1
)
(D
1
)
(D
1
D
2
)
[d E(D
1
D
2
)]
Using the independence of D
1
and D
2
, we have
E(D
1
D
2
) =
1

2
, (D
1
D
2
) =
_

2
1
+
2
2
and
cov(D
1
D
2
, D
1
) = cov(D
1
, D
1
) cov(D
1
, D
2
) =
2
1
0.
Hence the linear least square estimate of D
1
is

1
+

2
1

2
1
+
2
2
(d
1
+
2
).
Chapter 12
12E-1 Any linear combination of the random variables X and Y is of the
form aZ
1
+bZ
2
for constants a and b. Since the normally distributed random
variables Z
1
and Z
2
are independent, the random variable aZ
1
+ bZ
2
has
a normal distribution for any constants a and b. This is a basic result for
the normal distribution (see Section 10.4.7 in the book). Hence any linear
combination of the random variables X and Y is normally distributed. It
now follows from Rule 12.3 in the book that the random vector (X, Y ) has
a bivariate normal distribution. The parameters (
1
,
2
,
2
1
,
2
2
, ) of this
distribution are easily found. We have
1
= E(X) = 0,
2
= E(Y ) = 0,

2
1
=
2
(X) = 10,
2
2
=
2
(Y ) = 2. Since cov(X, Y ) = E(XY ) = E(Z
2
1
+
4Z
1
Z
2
+ 4Z
2
2
) = 1 + 4 = 5, it follows that = 5/

20 =
1
2

5.
12E-2 Since V and W are linear combinations of X and Y , the random vector
(V, W) has a bivariate normal distribution. The random variables V and W
are independent if and only if cov(V, W) = 0. It holds that
cov(V, W) = cov(aX, X) + cov(aX, aY ) + cov(Y, X) + cov(Y, aY )
= a 0.5a
2
0.5 +a = 0.5a
2
+ 2a 0.5.
The solutions of the equation 0.5a
2
+2a 0.5 = 0 are given by a = 2 +

3
and a = 2

3. The random variables V and W are independent for these


two values of a.
12E-3 Dene the random variables V and W by V = X and W = X +
_
1
2
Z. Any linear combination of V and W is a linear combination
80
of X and Z and is normally distributed, using the fact that X and Z are
independent and normally distributed. has a normal distribution. Hence the
random vector (V, W) has a bivariate normal distribution. The parameters
of this bivariate normal distribution are given by E(V ) = 0, E(W) = 0,

2
(V ) = 1,
2
(W) =
2
+ 1
2
= 1, and (V, W) = E(V W) = E
_
X
2
+
_
1
2
ZX
_
= . Hence the random vector (V, W) has the same standard
bivariate normal distribution as (X, Y ).
12E-4 Since P(X > 0) =
1
2
, we have P(Y > X | X > 0) = 2P(Y > X, X >
0). Next we use the result that
_
X, X +
_
1
2
Z
_
has the same bivariate
normal distribution as (X, Y ), where Z is an N(0, 1) distributed random
variable that is independent of X, see Problem 12E-3. Hence
P(Y > X | X > 0) = 2P
_
Z >
1
_
1
2
X, X > 0
_
.
This gives
P(Y > X | X > 0) =
2
2
_

0
e

1
2
x
2
dx
_

(1)x/

1
2
e

1
2
z
2
dz.
Using polar coordinates, we nd
_

0
e

1
2
x
2
dx
_

ax
e

1
2
z
2
dz =
_ 1
2

arctg(a)
_

0
e

1
2
r
2
r dr d
=

2
arctg(a)
for any constant a 0. This gives
P(Y > X | X > 0) =
1
2

1

arctg
_
1
_
1
2
_
.
(b) Write P(Y/X 1) as
P
_
Y
X
1
_
= P(Y X, X > 0) +P(Y X, X < 0).
Using the result of Problem 12E-3, it next follows that
P
_
Y
X
1
_
= P
_
Z
1
_
1
2
X, X > 0
_
+P
_
Z
1
_
1
2
X, X < 0
_
.
81
counts observed expected
0 57 54.3768
1 203 210.4604
2 383 407.2829
3 525 525.4491
4 532 508.4244
5 408 393.5610
6 273 253.8730
7 139 140.3699
8 45 67.9110
9 27 29.2046
10 10 11.3034
11 6 5.7831
Hence
P
_
Y
X
1
_
=
1
2
_

0
e

1
2
x
2
dx
_
(1)x/

1
2

1
2
z
2
dz
+
1
2
_
0

1
2
x
2
dx
_

(1)x/

1
2
e

1
2
z
2
dz
Using the method of polar coordinates to evaluate the integrals, we next nd
P
_
Y
X
1
_
=
1
2
+
1

arctg
_
1
_
1
2
_
.
12E-5 By the denition of the multivariate distribution, the random variables
a
1
X
1
+ + a
n
X
n
and b
1
Y
1
+ + b
m
Y
m
are normally distributed for any
constants a
1
, . . . , a
n
and b
1
, . . . , b
m
. The sum of two independent normally
distributed random variables is again normally distributed. Hence a
1
X
1
+
+ a
n
X
n
+ b
1
Y
1
+ + b
m
Y
m
is normally distributed for any constants
a
1
, . . . , a
n
and b
1
, . . . , b
m
, showing that (X, Y) has a multivariate normal
distribution.
12E-6 The value of the chi-square statistic is
(60.179 61, 419.5)
2
61, 419.5
+ +
(61, 334 61, 419.5)
2
61, 419.5
= 642.46
The chi-square distribution with 12-1=11 degrees of freedom is used in the
test. The probability
2
11
- distributed random variable takes on a value larger
82
than 642.46 is practically zero. This leaves no room at all for doubt about
the fact that birth dates are not uniformly distributed over the year.
12E-7 The parameter of the hypothesized Poisson distribution describing the
number of counts per time interval is estimated as
10,094
2,608
= 3.8704. We form
12 groups of time intervals by taking together the time intervals with 11 or
more particles. In the table we give the expected Poisson frequencies. The
value of the chi-square statistic is given by
(57 54.768)
2
54.3768
+
(203 210.4604)
2
210.4604
+ +
(6 5.7831)
2
5.7831
= 11.613.
Since the parameter of the hypothesized Poisson distribution has been esti-
mated from the data, the chi-square distribution with 1211 = 10 degrees
of freedom is used for the test. The probability P(
2
10
11.613) = 0.3117.
The Poisson distribution gives a good t.
12E-8 The parameter of the hypothesized Poisson distribution is estimated
as
=
37
78
vacancies per year. The data are divided in three groups: years with 0
vacancies, with 1 vacancy and with 2 vacancies. Letting p
i
= e

i
/i!, the
expected number of years with 0 vacancies is 78p
0
= 48.5381, with 1 vacancy
is 78p
1
= 23.0245 and with 2 vacancies is 78(1 p
0
p
1
) = 6.4374. The
chi-square test statistic with 3 1 1 = 1 degree of freedom has the value
(48 48.5381)
2
48.5381
+
(23 23.0245)
2
23.0245
+
(7 6.4374)
2
6.4374
= 0.055.
The probability P(
2
1
> 0.055) = 0.8145, showing that the Poisson distri-
bution with expected value =
37
78
gives an excellent t for the probability
distribution of the number of vacancies per year.
Chapter 13
13E-1 Let X be equal to 1 if the stronger team is the overall winner and let
X be equal to 0 otherwise. The random variable Y is dened as the number
of games the nal will take. The joint probability mass function of X and Y
satises
P(X = 0, Y = k) =
_
k 1
3
_
(0.45)
4
(0.55)
k4
for k = 4, . . . , 7
83
It now follows that the conditional probability mass function of Y given that
X = 0 is given by
P(Y = k, X = 0)
P(X = 0)
=
_
k1
3
_
(0.45)
4
(0.55)
k4

7
j=4
_
j1
3
_
(0.45)
4
(0.55)
j4
.
This conditional probability has the numerical values 0.1047, 0.2303, 0.3167,
and 0.3483 for k = 4, 5, 6, and 7.
13E-2 First the marginal distribution of X must be determined. Using the
binomium of Newton, it follows that
P(X = n) =
n

k=0
_
n
k
__
1
6
_
k
_
1
3
_
nk
=
_
1
3
+
1
6
_
n
, n = 1, 2, . . . .
That is, the random variable X is geometrically distributed with parameter
1
2
. Hence the conditional distribution of Y given that X = n is given by
P(Y = k | X = n) =
_
n
k
_ _
1
6
_
k
_
1
3
_
nk
_
1
2
_
n
=
_
n
k
__
1
3
_
n
2
k
, k = 0, 1, . . . , n.
Remark. Using the identity

n=k
_
n
k
_
a
nk
= (1 a)
k1
for |a| < 1, it is
readily veried that the marginal distribution of Y is given by P(Y = 0) =
1
2
and P(Y = k) =
3
2
_
1
4
_
k
for k = 1, 2, . . .. For k = 0 the conditional
distribution of X given that Y = k is P(X = n | Y = 0) = 2
_
1
3
_
n
for
n 1, while for k 1 the conditional distribution of X given that Y = k is
P(X = n | Y = k) =
_
n
k
_ _
2
3
_
k+1
_
1
3
_
nk
for n k.
13E-3 The conditional expected value E(X | Y = 2) is computed from
E(X | Y = 2) = 1 P(X = 1 | Y = 2) +

i=3
i P(X = i | Y = 2).
It holds that P(X = 1, Y = 2) =
1
6

1
6
and
P(X = i, Y = 2) =
4
6

1
6

_
5
6
_
i3

1
6
for i 3
Also, P(Y = 2) =
5
6

1
6
. Hence, by P(X = i | Y = 2) =
P(X=i, Y =2)
P(Y =2)
,
P(X = 1 | Y = 2) =
1
5
, P(X = i | Y = 2) =
4
30
_
5
6
_
i3
for i 3.
84
This leads to
E(X | Y = 2) =
1
5
+

i=3
i
4
30
_
5
6
_
i3
= 6.6.
In a similar way E(X | Y = 20) is computed. Using the expressions
P(X = i, Y = 20) =
_
4
6
_
i1
1
6
_
5
6
_
19i
1
6
for 1 i 19
P(X = i, Y = 20) =
_
4
6
_
19
1
6
_
5
6
_
i21
1
6
for i 21,
we nd that E(X | Y = 20) = 5.029.
13E-4 By f(x, y) = f
X
(x)f
Y
(y | x), it follows that
f(x, y) = 2 for 0 < x < 1, 0 < y < x
and f(x, y) = 0 otherwise. This leads to the marginal density f
Y
(y) =
_
1
y
f(x, y) dx = 2(1y) for 0 < y < 1. By f(x, y) = f
Y
(y)f
X
(x | y), we next
nd for any y with 0 < y < 1 that
f
X
(x | y) =
1
1 y
for y < x < 1
and f
X
(x | y) = 0 otherwise. This leads to
E(X | Y = y) =
_
1
y
x
1 y
dx =
1
2(1 y)
(1 y
2
) =
1
2
(1 +y), 0 < y < 1.
13E-5 The approach is to use the relation f(x, y) = f
X
(x)f
Y
(y | x) and
to simulate rst a random observation for x from f
X
(x) and to simulate
next a random observation for y from f
Y
(y | x). It follows from f
X
(x) =
_

x
f(x, y)dy that
f
X
(x) = e
x
for x > 0.
Next, by f
Y
(y | x) = f(x, y)/f
X
(x), it holds for any x that
f
Y
(y | x) = e
(yx)
for y > x.
Hence the marginal density f
X
(x) is the exponential density with parameter
1, while, for xed x, the conditional density f
Y
(y | x) is the exponential
density with parameter 1 shifted to the point x. A random observation
from the exponential density can be obtained by the inverse-transformation
85
method. Hence a random observation from f(x, y) can be simulated as
follows:
Step 1. Generate two random numbers u
1
and u
2
.
Step 2. Output x = ln(u
1
) and y = x ln(u
2
).
13E-6 The marginal density of the lifetime X of the rst circuit is given by
f
X
(x) =
_

1
24
(x +y)
4
dx =
8
(x + 1)
3
for x > 1
and f
X
(x) = 0 otherwise. Hence the conditional density of the lifetime of
the second circuit given that the rst circuit has failed after s time units is
given by
f
Y
(y | s) =
f(s, y)
f
X
(s)
=
3(s + 1)
3
(s +y)
4
for y > 1
and f
Y
(y | s) = 0 otherwise. Hence the expected value of the lifetime of the
second circuit given that the rst circuit has failed after s time units is given
by
E(Y | X = s) =
_

1
y
3(s + 1)
3
(s +y)
4
dy = (s + 1)
3
_

1
y d(s +y)
3
= 1 + (s + 1)
3
_

1
(s +y)
3
dy = s + 2
The probability that the second circuit will work more than v time units
given that the rst circuit has failed after s time units is equal to
_

v
f
Y
(y | s) dy =
(s + 1)
3
(s +v)
3
for v 1.
13E-7 Suppose you mark n boxes. Let the random variable X be your gain.
To nd E(X), condition upon the random variable Y which is dened to be
1 if the box with the devils penny is not among the n marked cards and is
0 otherwise. Then
E(X) = E(X | Y = 0)P(Y = 0) +E(X | Y = 1)P(Y = 1)
=
_
1
n
11
_
E(X | Y = 1).
The conditional expectation E(X | Y = 1) is equal to E(V
1
+ +V
n
), where
each V
i
is distributed as the dollar value of a randomly chosen box from the
boxes containing dollars. Hence E(V
i
) =
1
10

10
i=1
a
i
. Letting the constant
v =
1
10

10
i=1
a
i
and noting that E(V
1
+ + V
n
) = E(V
1
) + + E(V
n
), it
follows that
E(X) =
_
1
n
11
_
vn.
86
The function x
x
2
11
of the continuous variable x is maximal for x = 5.5.
Hence the best choice for n is 5 or 6. The expression x
x
2
11
has for n = 6 the
same same value as for n = 5. Hence the optimal decision is to mark 5 or 6
boxes. This decision results in an expected payo of
3
11

10
i=1
a
i
dollars. Note:
It is interesting to compare the solution of Problem 13E-4 to the solution of
Problem 9E-6.
13E-8 Let the random variable L be the length of the subinterval covering
the number s. By the law of conditional expectation,
E(L) =
_
1
0
E(L | U = u)f
U
(u) du =
_
1
0
E(L | U = u)du.
This gives
E(L) =
_
s
0
(1 u) du +
_
1
s
udu = s s
2
+
1
2
.
13E-9 (a) Let the random variable Y be the number of heads showing up
in the rst ve tosses of the coin. Then, E(X | Y = y) = y +
1
2
y for
0 y 5. Hence, by E(X) =

5
y=0
E(X | Y = y)P(Y = y), we have
E(X) = E(Y )+
1
2
E(Y ). Noting that E(Y ) =
5
2
, it follows that E(X) =
15
4
=
3.75.
(b) Let the random variable Y be the outcome of the rst roll of the die.
Then, E(X | Y = y) = (6 y) +
1
6
y for 0 y 6, where (0) = 1 and
(k) = 0 for k = 0. Hence, by E(X) =

6
y=0
E(X | Y = y)P(Y = y), we
have E(X) =
1
6
E(Y ) +
1
6
1. Since E(Y ) = 3.5, it follows that E(X) =
4.5
6
= 0.75.
13E-10 By the convolution formula,
P(X
1
+X
2
= r) =
r1

k=1
p(1 p)
k1
p(1 p)
rk1
.
This gives
P(X
1
= j | X
1
+X
2
= r) =
P(X
1
= j, X
1
+X
2
= r)
P(X
1
+X
2
= r)
=
p(1 p)
j1
p(1 p)
rj1

r1
k=1
p(1 p)
k1
p(1 p)
rk1
=
1
r 1
.
In other words, the conditional distribution of X
1
given that X
1
+X
2
= r is
the discrete uniform distribution on 1, . . . , r 1.
87
13E-11 The number p is a random observation from a random variable U
that is uniformly distributed on (0, 1). Let the random variable X denote
the number of times that heads will appear in n tosses of the coin. Then, by
the law of conditional probability,
P(X = k) =
_
1
0
P(X = k | U = p) dp =
_
1
0
_
n
k
_
p
k
(1 p)
nk
dp
for k = 0, 1, . . . , n. Using the result for the beta integral, we next obtain
P(X = k) =
_
n
k
_
k!(n k)!
(n + 1)!
=
1
n + 1
for k = 0, 1, . . . , n.
13E-12 Let V = X
1
and W = X
1
+X
2
. To nd the joint density function of
V and W, we apply the transformation rule from Section 11.4. The inverse
functions a(v, w) and b(v, w) are given by a(v, w) = v and b(v, w) = w v.
The Jacobian J(v, w) = 1. Hence the joint density function of V and W is
given by
f
V,W
(v, w) = e
v
e
(wv)
=
2
e
w
for 0 < v < w
and f
V,W
(v, w) = 0 otherwise. The density function of W is given by
_
0
w

2
e
w
dv =
2
we
w
for w > 0 (the Erlang-2 density with scale pa-
rameter ). Hence the conditional density function of V given that W = s
is equal to
f
V
(v | s) =

2
e
s

2
se
s
=
1
s
for 0 < v < s
and f
V
(v | s) = 0 otherwise. In other words, the conditional probability
density function of X
1
given that X
1
+ X
2
= s is the uniform density on
(0, s).
13E-13 The joint density function f(, r) of and R is given by f(, r) =
1
2
re

1
2
r
2
for < < and r > 0. To nd the joint density function of
V and W, we apply the transformation rule from Section 11.4. The inverse
functions = a(v, w) and r = b(v, w) are given by a(v, w) = arctg(w/v) and
b(v, w) =

v
2
+w
2
. Using the fact that the derivative of arctg(x) is equal
to 1/(1 + x
2
), we nd that the Jacobian J(v, w) =

v
2
+w
2
. Hence the
joint density function of V and W is equal to
f
V,W
(v, w) =
1
2
(v
2
+w
2
)e

1
2
(v
2
+w
2
)
for < v, w < .
Using the fact that
1

2
_

1
2
v
2
dv = 1 and
1

2
_

v
2
e

1
2
v
2
dv = 1, it
follows that the marginal density function of W is given by
f
W
(w) =
1

2
(1 +w
2
)e

1
2
w
2
for < w < .
88
The random variable V has the same marginal density as W. The conditional
density function of V given that W = w is equal to
f
V
(v | w) =
1

2
v
2
+w
2
1 +w
2
e

1
2
v
2
for < v < .
Since
_

v
k
e

1
2
v
2
dv = 0 for both k = 1 and k = 3, it follows that
E(V | W = w) =
1

2
_

v
3
+vw
2
1 +w
2
e

1
2
v
2
dv = 0 for all w.
13E-14 By the law of conditional probability,
P(N = k) =
_
1
0
P(N = k | X
1
= u) du for k = 2, 3, . . . .
The conditional probability distribution of N given that X
1
= u is a shifted
geometric distribution with success parameter p = 1 u. Hence
P(N = k) =
_
1
0
u
k2
(1 u)du =
1
k(k 1)
for k = 2, 3, . . . .
The expected value of N is given by E(N) = .
13E-15 For any 0 x 1,
P(U
1
x | U
1
> U
2
) =
P(U
1
x, U
1
> U
2
)
P(U
1
> U
2
)
=
_
x
0
du
1
_
u
1
0
du
2
1/2
= x
2
and
P(U
2
x | U
1
> U
2
) =
P(U
2
x, U
1
> U
2
)
P(U
1
> U
2
)
=
_
x
0
du
2
_
1
u
2
du
1
1/2
= 2
_
x
1
2
x
2
_
.
Thus the conditional densities of U
1
and U
2
given that U
1
> U
2
are dened
by 2x and 2(1 x) for 0 < x < 1 and are zero otherwise. This gives
E(U
1
| U
1
> U
2
) =
_
1
0
x 2xdx =
2
3
, E(U
2
| U
1
> U
2
) =
_
1
0
x 2(1x) dx =
1
3
.
13E-16 The two roots of the equation x
2
+ 2Bx + 1 = 0 are real only if
|B| 1. Since P(|B| 1) = P(B 1) +P(B 1), the probability of two
real roots is equal to
P(|B| 1) = (1) + 1 (1) = 2[1 (1)] = 0.3174.
89
The sum of the two roots is 2B. To determine the conditional density
function of 2B given that |B| 1, note that P(2B x

|B| 1) is given
by
P(2B x

|B| 1) = P(B
1
2
x

|B| 1) =
P(B
1
2
x, |B| 1)
P(|B| 1)
.
We have P(B
1
2
x, |B| 1) = P(B 1) for 2 < x < 2. Further
P(B
1
2
x, |B| 1) =
_
P(B
1
2
x) = 1 (
1
2
x), x 2
P(
1
2
x B 1) = (1) (
1
2
x), x 2.
Denoting by (x) =
1

2
e

1
2
x
2
the standard normal density, it now follows
that the conditional probability density function of 2B given that |B| 1
is equal to
1
4
(
1
2
x)/[1 (1)] for both x < 2 and x > 2 and is equal to 0
for 2 x 2.
13E-17 Let the random variable T be the transmission time of a message and
N be the number of newly arriving messages during the time T. Under the
condition that T = n, the random variable N has a binomial distribution
with parameters n and p. Hence, by the law of conditional expectation,
E(N) =

n=1
E(N | T = n)a(1 a)
n1
and
E(N
2
) =

n=1
E(N
2
| T = n)a(1 a)
n1
.
The rst two moments of a binomial distribution with parameters n and p are
given by np and np(1p) +n
2
p
2
, while the rst two moments of a geometric
distribution with parameter a are given by
1
a
and
1a
a
2
+
1
a
2
. This gives
E(N) =

n=1
npa(1 a)
n1
=
p
a
and
E(N
2
) =

n=1
_
np(1 p) +n
2
p
2
_
a(1 a)
n1
=
p(1 p)
a
+
p
2
(2 a)
a
2
.
90
Hence the expected value and the standard deviation of the number of newly
arriving messages during the transmission time of a message are given by p/a
and
_
p[a(1 p) +p(1 a)]/a.
13E-18 Let the random variable Y denote the number of messages waiting in
the buer. Then P(Y = y) = p(1p)
y1
for y 1. By the law of conditional
probability,
P(X = x) =

y=x+1
P(X = x | Y = y)p(1 p)
y1
=

y=x+1
1
y
p(1 p)
y1
for x = 0, 1, . . .. Using the formula

n=1
u
n
n!
= ln(1 u) for |u| < 1, the
expression for P(X = x) can be written as
P(X = x) =
p
1 p
ln(p)
x

y=1
1
y
p(1 p)
y1
.
Using the fact that the discrete uniform distribution on 0, 1, . . . , y 1 has
expected value
1
2
(y 1), the expected value of X is calculated from
E(X) =

y=1
E(X | Y = y)p(1p)
y1
=

y=1
1
2
(y1) p(1p)
y1
=
1
2
_
1
p
1
_
.
13E-19 Given that the carnival master tells you that the ball picked from
the red beaker has value r, let L(r) be your expected payo when you guess
a larger value and let S(r) your expected payo when you guess a smaller
value. Then
L(r) =
1
10
10

k=r+1
k +
r/2
10
=
1
20
(10 r)(r + 11) +
r/2
10
=
110 r
2
20
and
S(r) =
1
10
r1

k=1
k +
r/2
10
=
1
20
(r 1)r +
r/2
10
=
r
2
20
.
It holds that L(r) > S(r) for 1 r 7 and L(r) < S(r) for 8 r 10,
as follows by noting that 110 x
2
= x
2
has x

55 7.4 as solution.
Thus, given that the carnival master tells you that the ball picked from the
red beaker has value r, your expected payo is maximal by guessing a larger
value if r 7 and guessing a smaller value otherwise. Applying the law of
conditional expectation. it now follows that your expected payo is
7

k=1
110 r
2
20

1
10
+
10

k=8
r
2
20

1
10
= 4.375 dollars
91
if you use the decision rule with critical level 7. The game is not fair, but the
odds are only slightly in favor of the carnival master if you play optimally.
Then the house edge is 2.8% (for critical levels 5 and 6 the house edge has
the values 8.3% and 4.1%).
13E-20 For the case that you start with n strings, let X
n
be the number of
loops you get and let Y
n
= 1 if the rst two loose ends you choose are part of
the same string and Y
n
= 0 otherwise. By the law of conditional expectation,
E(X
n
) = E(X
n
| Y
n
= 1)P(Y
n
= 1) +E(X
n
| Y
n
= 0)P(Y
n
= 0) and so
E(X
n
) =
_
1 +E(X
n1
)

1
2n 1
+E(X
n1
)
_
1
1
2n 1
_
=
1
2n 1
+E(X
n1
) for n = 1, 2, . . . , N,
where E(X
0
) = 0. This gives E(X
n
) =

2n1
i=1
1
i
for n = 1, . . . , N.
13E-21 (a) Denote by N the Poisson distributed random variable you sam-
ple from. Let X be the number of heads you will obtain. By the law of
conditional probability, for any k = 0, 1, . . .,
P(X = k) =

n=k
P(X = k | N = n)P(N = n) =

n=k
_
n
k
__
1
2
_
n
e
1
n!
=
e
1
(1/2)
k
k!

n=k
(1/2)
nk
(n k)!
=
e
1
(1/2)
k
k!
e
1
2
=
e

1
2
(1/2)
k
k!
.
In other words, the number of heads you will obtain is Poisson distributed
with an expected value of 0.5.
(b) Let the random variable U be uniformly distributed on (0, 1). Then, by
conditioning on U,
P(X = k) =
_
1
0
P(X = k | U = u) du =
_
1
0
e
u
u
k
k!
du.
This integral can be interpreted as the probability that an Erlang-distributed
random variable with shape parameter k +1 and scale parameter 1 will take
on a value smaller than or equal to 1. This shows that
P(X = k) =

j=k+1
e
1
1
j!
for k = 0, 1, . . . .
92
13E-22 By the law of conditional probability,
P(N = k) =
_

0
e
x
x
k
k!
x
r1
[(1 p)/p]
r
(r)
e
(1p)x/p
dx
=
(1 p)
r
p
r
k! (r)
_

0
x
r+k1
e
x/p
dx.
Since the gamma density x
r+k1
(1/p)
r+k
e
x/p
/(r + k) integrates to 1 over
(0, ), it next follows that
P(N = k) =
(r +k)
k! (r)
p
k
(1 p)
r
for k = 0, 1, . . . .
In other words, the random variable has a negative binomial distribution,
where the scale parameter r is not necessarily integer-valued.
13E-23 The random variable Z has the probability density function f(x) =
_
2

1
2
x
2
for x > 0. The constant c such that f(x) cg(x) for all x is easily
found. Since f(x)/g(x) =
_
2/ e
x
1
2
x
2
, it follows by dierentiation that
f(x)/g(x) is maximal at x = 1 and has
_
2e/ as its maximal value. Hence
c =
_
2e/ ( 1.32). Note that f(x)/cg(x) = e

1
2
(x1)
2
. The acceptance-
rejection method proceeds as follows:
Step 1. Generate two random numbers u
1
and u
2
. Let y = ln(u
1
) be the
random observation from the exponential density with parameter 1.
Step 2. If u
2
e

1
2
(y1)
2
, then v = ln(u
1
) is a random observation for the
random variable |Z|. Otherwise, return to Step 1.
Suppose that v is a random observation for the random variable |Z|.
Then a random observation z from the standard normal density follows by
generating a random number u and taking z = v if u 0.5 and z = v
otherwise.
13E-24 Let the random variable represent the unknown probability that
a single toss of the die results in the outcome 6. The prior density of
is given by f
0
) =
10
3
for (0, 1, 0.4). The posterior density f( | data)
is proportional to L(data | )f
0
(), where L(data | ) =
_
300
75
_

75
(1 )
225
.
Hence the posterior density is given by
f( | data =

75
(1 )
225
_
0.4
0.1
x
75
(1 x)
225
dx
for 0.1 < < 0.4.
and f( | data = 0 otherwise. The posterior density is maximal at =
75
75+225
= 0.25. The solution of
_
x
0
f( | data) d = is given by 0.2044 for
93
= 0.025 and by 0.3020 for = 0.975. Hence a 95% Bayesian condence
interval for is given by (0.2044, 0.3020).
13E-25 Let the random variable represent the unknown probability that
a free throw of your friend will be successful. The posterior density function
f( | data) is proportional to L(data | )f
0
(), where L(data | ) =
_
10
7
_

7
(1
)
3
. Hence the posterior density is given by
f( | data) =
_
_
_
c
1

7
(1 )
3
16( 0.25) for 0.25 < < 0.50,
c
1

7
(1 )
3
16(0.75 ) for 0.50 < < 0.75,
where
c = 16
_
0.50
0.25
x
7
(1 x)
3
(x 0.25)dx + 16
_
0.75
0.50
x
7
(1 x)
3
(0.75 x)dx.
The posterior density f( | data) is zero outside the interval (0.25, 0.75).
Numerical calculations show that f( | data) is maximal at = 0.5663.
Solving
_
x
0.25
f( | data)d = for = 0.025 and 0.975 leads to the 95%
Bayesian condence interval (0.3883, 0.7113).
13E-26 The likelihood L(data | ) is given by
L(data | ) =
1

2
e

1
2
(t
1
)
2
/
2
1
,
where t
1
= 123 is the test score. Hence the posterior density f( | data) is
proportional to
e

1
2
(t
1
)
2
/
2
1
e

1
2
(
0
)
2
/
2
0
.
Next a little algebra shows that f( | data) is proportional to
e

1
2
()
2
/
2
,
where
=

2
0
t
1
+
2
1

2
0
+
2
1
= 118.4 and
2
=

2
0

2
1

2
0
+
2
1
= 45.
In other words, the posterior density is a normal density with an expected
value of 118.4 and a standard deviation of 6.708. The posterior density
is maximal at = 118.4. Using the 0.025 and 0.975 percentiles -1.960 and
1.960 of the standard normal distribution, a Bayesian 95% condence interval
for the true value of the IQ is calculated as ( 1.960, + 1.960) =
(105.3, 131.5).
94
Chapter 14
14E-1 Put for abbreviation p
n
= P(X = n). By the denition of G
X
(z), we
have G
X
(1) =

n=0
p
2n

n=0
p
2n+1
. Also,

n=0
p
2n
+

n=0
p
2n
= 1.
Hence G
X
(1) + 1 = 2

n=0
p
2n
, showing the desired result.
14E-2 Let the random variable X be the outcome of the rst roll of the die.
By conditioning on X,
E(z
S
) =
6

j=1
1
6
E(z
S
| X = j) =
1
6
6

j=1
E(z
j+X
1
++X
j
),
where X
1
, . . . , X
j
are independent random variables each having the discrete
uniform distribution on {1, 2, . . . , 6}. This gives
E(z
S
) =
1
6
6

j=1
z
j
_
1
6
z +
1
6
z
2
+ +
1
6
z
6
_
j
=
1
6
6

j=1
_
1
6
z
2
+
1
6
z
3
+ +
1
6
z
7
_
j
.
Using the rst two derivatives of E(z
S
), we nd E(S) = 15.75 and E(S
2
) =
301.583. Hence the expected value and the standard deviation of the sum S
are given by 15.75 and

301.583 15.75
2
=7.32.
14E-3 By conditioning on T and using the fact that e
m(1z)
is the generating
function of a Poisson distributed random variable with expected value m, it
follows that the generating function G
X
(z) = E(z
X
) is given by
G
X
(z) =
_

0
E(z
X
| T = t)e
t
dt =
_

0
e
t(1z)
e
t
dt
=

+(1 z)
.
This generating function is recognized as the generating function of the prob-
ability mass function {p(1p)
j
, j = 0, 1, . . .} with p = /(+). The gener-
ating function of a nonnegative, integer-valued random variable determines
uniquely its probability mass function. Hence the random variable X has a
geometric distribution on {0, 1, . . .} with parameter /( +).
14E-4 The random variable X is distributed as 1+X
1
with probability p and
distributed as 1+X
2
with probability 1p. By conditioning on the outcome
of the rst trial, it follows that E(z
X
) = pE(z
1+X
1
)+(1p)E(z
1+X
2
). Hence
E(z
X
) = pzE(z
X
1
) + (1 p)zE(z
X
2
).
95
The random variable X
1
is equal to 2 with probability p
2
, is distributed as
1 + X
2
with probability 1 p and is distributed as 2 + X
2
with probability
p(1 p). This gives
E(z
X
1
) = p
2
z
2
+ (1 p)zE(z
X
2
) +p(1 p)E(z
X
2
)
By the same argument
E(z
X
2
) = (1 p)
2
z
2
+pzE(z
X
1
) + (1 p)pE(z
X
1
).
Solving the two linear equations in E(z
X
1
) and E(z
X
2
) gives
E(z
X
1
) =
p
2
z
2
+ (1 p)
3
z
3
+p(1 p)
3
z
4
1 [(1 p)pz
2
+p(1 p)z
3
+p
2
(1 p)
2
z
4
]
.
By interchanging the roles of p and 1 p in the right-hand side of this
expression, the formula for E(z
X
2
) follows. We can now conclude that the
generating function of X is given by
E(z
X
) =
p
3
z
3
+p(1 p)
3
z
4
+p
2
(1 p)
3
z
5
1 [(1 p)pz
2
+p(1 p)z
3
+p
2
(1 p)
2
z
4
]
+
(1 p)
3
z
3
+ (1 p)p
3
z
4
+ (1 p)
2
p
3
z
5
1 [p(1 p)z
2
+ (1 p)pz
3
+ (1 p)
2
p
2
z
4
]
.
For the special case of p =
1
2
(fair coin), this expression can be simplied to
E(z
X
) =
(1/4)z
3
1z/2z
2
/4
.
Remark. By dierentiating E(z
X
) with respect to z and putting z = 1, we
nd after tedious algebra that
E(X) = A(p) +A(1 p),
where
A(x) =
x
9
5x
8
+ 6x
7
+ 5x
6
16x
5
+ 11x
4
+ 7x
3
10x
2
+ 4x
_
1 2x(1 x) (1 x)
2
x
2
_
2
.
As a check, E(X) = 7 for the special case of p =
1
2
. It is always a good plan
to check your answers, if possible, by reference to simple cases where you
know what the answers should be.
14E-5 The moment-generating function
M
X
(t) =
_

e
tx
e
x
(1 +e
x
)
2
dx
96
is dened only for 1 < t < 1. Using the change of variable u = 1/(1 + e
x
)
with
du
dx
= e
x
/(1 +e
x
)
2
, it follows that
M
X
(t) =
_
1
0
_
1 u
u
_
t
du for 1 < t < 1.
This integral is the well-known beta-integral
_
1
0
u
t
(1 u)
t
du that can be
evaluated as (1 +t)(1 t), where (x) is the gamma function. Using the
fact that the derivative of the function a
x
is ln(a)a
x
for any constant a > 0,
it is easily veried that
M

X
(0) =
_
1
0
ln
_
1 u
u
_
du and M

X
(0) =
_
1
0
ln
2
_
1 u
u
_
du.
The two integrals can be evaluated as
_
1
0
[ln(1 u) ln(u)]du = 0,
_
1
0
[2 ln
2
(u) 2 ln(u) ln(1 u)]du =

2
3
,
showing that E(X) = 0 and
2
(X) =

2
3
.
14E-6 The relation M
X
(t) = e
t
M
X
(t) can be written as E(e
tX
) = e
t
E(e
tX
) =
E(e
t(1X)
) and is thus equivalent with
M
X
(t) = M
1X
(t) for all t.
Since the moment-generating function determines uniquely the probability
distribution, it follows that the random variable X has the same distribution
as the random variable 1 X. Hence E(X) = E(1 X) and so E(X) =
1
2
.
The relation M
X
(t) = e
t
M
X
(t) is not sucient to determine the density of
X. The property is satised for both the uniform density on (0, 1) and the
beta density 6x(1 x) on (0, 1).
14E-7 (a) Chebyshevs inequality states that
P(|X | c)

2
c
2
for any constant c > 0.
Taking c = k, it follows that the random variable X falls in the interval
( k, + k) with a probability of at least 1 1/k
2
. The inequality
1 1/k
2
p leads to the choice k = 1/

1 p.
(b) For any constant a > 0,
P(X > +a) P(|X| a) P(|X| a+)

2
(a +)
2


2
a
2
+
2
,
97
where the third inequality uses Chebyshevs inequality.
14E-8 (a) The moment-generating function of X is given by
M
X
(t) =
_
1
1
e
tx
1
2
dx = t
1
_
e
t
e
t
2
_
for < t < .
The function
1
2
(e
t
e
t
) is also known as sinh(t) and has the power series
representation t +
t
3
3!
+
t
5
5!
+ .
(b) Put for abbreviation X
n
=
1
n
(X
1
+ +X
n
). Using the assumption that
the X
i
are independent, it follows that the generating function of X
n
satises
M
X
n
(t) = e
t
n
(X
1
++X
n
)
= e
t
n
X
1
e
t
n
X
n
.
Using the result in (a), we get
M
X
n
(t) =
_
t
n
_
n
_
e
t/n
e
t/n
2
_
n
.
This gives
P(X
n
c) = min
t>0
e
ct
_
t
n
_
n
_
e
t/n
e
t/n
2
_
n
.
Next we use the inequality
1
2
(e
t
e
t
) = t +
t
3
3!
+
t
5
5!
+ te
t
2
/6
for t > 0.
Hence
P(X
n
c) = min
t>0
e
ct
e
t
2
/6n
.
The function e
(ctt
2
/6n)
takes on its minimal value for t = 3cn. This gives
the desired bound P(X
n
c) e

3
2
cn
2
for c > 0.
14E-9 The moment generating function M
X
(t) = E(e
tX
) is given by
M
X
(t) =

n=0
e
tn
e

n
n!
= e
(1e
t
)
, < t < .
To work out P(X c) min
t>0
e
ct
M
X
(t), consider the function a(t) =
ct +(1 e
t
) for t 0. Since c > , this function is increasing at t = 0 with
a(0) = 0. Further a(t) tends to as t gets large. Putting the derivative
of a(t) equal to zero, we nd t = ln(c/). We can now conclude that the
98
function e
ct
M
X
(t) on (0, ) takes on its absolute minimum at the point
t = ln(c/) > 0 provided that c > . This gives the Cherno bound
P(X c)
_

c
_
c
e
c
.
14E-10 The extinction probability is the smallest root of the equation u =
P(u), where the generating function P(u) is given by
P(u) =

k=0
p(1 p)
k
u
k
=
p
1 (1 p)u
for |u| 1.
Writing the equation u =
p
1(1p)u
as (1 p)u
2
u + p = 0, it follows that
the equation u = P(u) has the two roots u =
p
1p
and u = 1. Hence the
extinction probability is
p
1p
if p <
1
2
and is 1 otherwise.
14E-11 Let X(t) be the number of orders present at time t. The continuous-
time process {X(t)} regenerates itself each time a production run is started.
We take as cycle the time between two successive production runs. The
expected length of one cycle is the expected amount of time needed to collect
N orders and so
the expected length of one cycle = + + = N.
The ih arriving order in a cycle is kept in stock until N i additional orders
have been arrived and so the expected holding cost incurred for this order is
h (N i). This gives
the expected costs incurred in one cycle = K +h(N 1) + +h
= K +
1
2
hN(N 1).
By dividing the expected costs incurred in one cycle by the expected length
of one cycle, it follows from the renewal-reward theorem that
the long-run average cost per unit time =
K
N
+
1
2
h(N 1)
with probability 1. The function g(x) = K/(x) +
1
2
hx is convex in x > 0
and has an absolute minimum in x

=
_
2K/(h). Hence the optimal value
of N is one of the integers nearest to x

.
14E-12 Let X(t) be the number of messages in the buer at time t. The
stochastic process {X(t), t 0} is regenerative. By the memoryless property
99
of the Poisson arrival process, the epochs at which the buer is emptied are
regeneration epochs. Let a cycle be the time between two successive epochs
at which the buer is emptied. Obviously, the length of a cycle is equal to
T. Let the random variable
n
denote the arrival time of the nth message.
Dene the cost function g(x) by g(x) = h (T x) for 0 x T and
g(x) = 0 otherwise. Then, the expected value of the total holding cost in the
rst cycle is
E
_

n=1
g(
n
)
_
= h

n=1
E[g(
n
)] = h

n=1
_
T
0
(T x)f
n
(x) dx,
where f
n
(x) is the probability density function of the arrival epoch
n
. The
interchange of the order of expectation and summation is justied by the
non-negativity of the function g(x). Since the interarrival times of messages
are independent random variables having an exponential distribution with
parameter , the density f
n
(x) is the Erlang density
n
x
n1
e
x
/(n 1)!.
Interchanging the order of summation and integration in the above expression
for the expected holding cost and noting that

n=1
f
n
(x) = , it follows that
the expected value of the total holding cost in the rst cycle is equal to
h
_
T
0
(T x)

n=1
f
n
(x) dx = h
_
T
0
(T x)dx =
1
2
hT
2
.
Hence the total expected cost in one cycle is K+h
1
2
T
2
and so, by the renewal
reward theorem,
the long-run average cost per unit time =
K +
1
2
hT
2
T
=
K
T
+
1
2
hT
with probability 1. Putting the derivative of this convex cost function equal
to zero, it follows that the optimal value of T is given by
T

=
_
2K
h
.
14E-13 Let X(t) be the number of passengers waiting for departure of the
boat at time t. The stochastic process {X(t), t 0} is regenerative. The
regeneration epochs are the departure epochs of the boat. Consider the rst
cycle (0, T). By conditioning on the arrival time of the nth potential customer
and using the law of conditional probability, we have that
P(the nth arrival joins the rst trip) =
_
T
0
e
(Tt)

n
t
n1
e
t
(n 1)!
dt.
100
It now follows that
E(number of passengers on the rst trip) =

n=1
_
T
0
e
(Tt)

n
t
n1
e
t
(n 1)!
dt.
Interchanging the order of summation and integration and using the fact that

n=1

n
t
n1
e
t
/(n 1)! = , we nd
E(number of passengers on the rst trip) =
_
T
0
e
(Tt)
dt =

(1 e
T
).
Hence
E(the expected net reward in one cycle) =
R

(1 e
T
) K.
The expected length of one cycle is T. Hence, by the renewal-reward theorem,
the long-average net reward per unit time =
(R/)(1 e
T
) K
T
with probability 1. Assuming that R/ > K, dierentiation gives that
the long-run average net reward is maximal for the unique solution of the
equation e
T
(RT +R/) = R/ K.
14E-14 Dene the following random variables. For any t > 0, dene the
indicator variable I(t) by
I(t) =
_
1 if the system is out of stock at time t
0 otherwise
Also, for any n = 1, 2, . . ., dene the indicator variable I
n
by
I
n
=
_
1 if the system is out of stock when the nth demand occurs
0 otherwise.
The continuous-time stochastic process {I(t)} and the discrete-time stochas-
tic process {I
n
} are both regenerative. The regeneration epochs are the
demand epochs at which the stock on hand drops to zero, by the memoryless
property of the Poisson process. A cycle starts each time the stock on hand
drops to zero. The system is out of stock during the time elapsed from the
beginning of a cycle until the next inventory replenishment. This amount
of time is exponentially distributed with mean 1/. The expected amount
101
of time it takes to go from stock level Q to 0 equals Q/. Hence, by the
renewal-reward theorem,
the long-run fraction of time the system is out of stock
=
1/
1/ +Q/
.
To nd the fraction of demand that is lost, note that the expected amount
of demand lost in one cycle equals E(amount of time the system is out
of stock during one cycle) = /. Hence, by the renewal-reward theorem,
the long-run fraction of demand that is lost
=
/
/ +Q
.
The above two results lead to the remarkable nding:
the long-run fraction of customers nding the system out of stock
= the long-run fraction of time the system is out of stock.
This nding is a particular instance of the property Poisson arrivals see time
averages, see also Chapter 16 of the textbook. Roughly stated, this property
expresses that in statistical equilibrium the distribution of the state of the
system just prior to an arrival epoch is the same as the distribution of the
state of the system at an arbitrary epoch when arrivals occur according to a
Poisson process. This result requires only that the arrival process {N(t), t
0} can be seen as an exogenous factor to the system and is not aected by the
system itself. An intuitive explanation of the property Poisson arrivals see
time averages is that Poisson arrivals occur completely randomly in time.
14E-15 Dene the following random variables. For any t 0, let
I(t) =
_
1 if the work station is busy at time t.
0 otherwise.
Also, for any n = 1, 2, . . ., let
I
n
=
_
1 if the work station is busy just prior to the nth arrival
0 otherwise.
The continuous-time process {I(t)} and the discrete-time process {I
n
} are
both regenerative. The arrival epochs occurring when the working station
is idle are regeneration epochs for the two processes Let us say that a cycle
starts each time an arriving job nds the working station idle. The long-run
102
fraction of time the working station is busy is equal to the expected amount
of time the working station is busy during one cycle divided by the expected
length of one cycle. The expected length of the busy period during one
cycle equals . The Poisson arrival process is memoryless and so the time
elapsed from the completion of a job until a next job arrives has the same
distributions as the interarrival times between the jobs. Hence the expected
length of the idle period during one cycle equals the mean interarrival time
1/. Hence, by the renewal-reward theorem,
the long-run fraction of time the work station is busy
=

+ 1/
.
The long-run fraction of jobs that are lost equals the expected number of
jobs lost during one cycle divided by the expected number of jobs arriving
during one cycle. The arrival process is a Poisson process and so the expected
number of arrivals in any time interval of the length x is x. This gives that
the expected number of lost arrivals during the busy period in one cycle
equals E(processing time of a job) = . Hence, by the renewal-reward
theorem,
the long-run fraction of jobs that are lost
=

1 +
.
The above two results show that
the long-run fraction of arrivals nding the work station busy
= the long-run fraction of time the work station is busy.
Again we nd that Poisson arrivals see time averages. Also, it is interesting
to observe that in this particular problem the long-run fraction of lost jobs
is insensitive to the form of the distribution function of the processing time
but needs only the rst moment of this distribution. Problem 14E-15 is a
special case of Erlangs loss model that has been discussed in Chapter 16.
103

You might also like