IEOR 4106: Spring 2011

Solutions to Homework Assignment 7: Due on Tuesday, March 22.

More of Chapter 5: Read the rest of Section 5.3, skipping Examples 5.17 (Coupon Collect-
ing), 5.21 (Insurance claims)and Subsection 5.3.6 (software reliability). Read Section 5.4 up to
(but not including) Example 5.27 (Busy periods in the single server queue). Poisson process
and nonhomogeneous Poisson process.

31. Let T be the amount of time (minutes) that the 1:30 pm appointment spends at
the doctor’s office. Let the event A ≡ {1 pm appointment leaves before 1:30pm} and the
event Ac ≡ {1 pm appointment leaves after 1:30pm}. Denote the service times of the 1pm
appointment and 1:30pm appointment by S1 and S2 , respectively. Then by assumption S1 and
1
S2 are iid exp( 30 ). By conditioning on whether the 1 pm appointment has left or not when
the 1:30 pm appointment arrives on time, we obtain
E[T ] = E[T |A]P (A) + E[T |Ac ]P (Ac )
= 30P (S1 ≤ 30) + (30 + 30)P (S1 > 30)
= 30(1 + P (S1 > 30))
1
= 30(1 + e− 30 30 ) = 30(1 + e−1 ).

34. Let TA and TB be the life time of the two individuals A and B, respectively. Let T1 be
the time until the first arrival of kidney and T2 be the time between the first arrival and the
second arrival of kidneys. Then TA ∼ exp(µA ) and TB ∼ exp(µB ), Ti ∼ exp(λ), i = 1, 2. All
these variables are mutually independent.
(a) The event that A obtains a new kidney happens when the first kidney arrives before A
dies, i.e. T1 < TA , so
λ
P (A obtains a new kidney) = P (T1 < TA ) = .
λ + µA
(b) To obtain the probability of B obtains a new kidney, we will condition on the first
event, i.e., whether the first kidney arrives before either A or B dies or whether A dies first,
or whether B dies first. First, we need not consider the probability that B dies first, because
clearly B does not get the kidney then.
P (B obtains a new kidney)
= P (B obtains a new kidney|T1 < TA , T1 < TB )P (T1 < TA , T1 < TB )
+ P (B obtains a new kidney|TA < T1 , TA < TB )P (TA < T1 , TA < TB )
= P (T2 < TB )P (T1 < TA , T1 < TB ) + P (T1 < TB )P (TA < T1 , TA < TB )
³ λ ´³ λ ´ ³ λ ´³ µA ´
= +
λ + µB λ + µA + µB λ + µB λ + µA + µB
³ λ ´³ λ + µ ´
A
=
λ + µB λ + µA + µB
Directly we get the last line, by first having either the kidney arrive or A die first and then,
afterwards (and thus multiplying) the kidney arrive before B dies.
 39. Recall Problem 4 (e) on the midterm exam. We can use a normal approximation by
virtue of the central limit theorem. See Section 2.7 if you need a refresher. By the assumptions,
the lifetime is the sum of 196 IID exponential random variables each with mean 1/2.5 years,
i.e.,

S196 = X1 + · · · + X196 ,

where EXi = 1/2.5. Thus,

(a) the mean is
E[S196 ] = 196/2.5 = 78.4 .

(b) and the variance is

V ar[S196 ] = 196(1/2.5)2 = 31.36 .

For parts (c) - (e), use the normal approximation, which follows from the central limit
theorem; see Section 2.7. Use the normal probabilities in Table 2.3. Let Z be a standard
normal random variable (with mean 0 and variance 1).
(c)
P (S196 < 67.2) ≈ P (Z < (67.2 − 78.4)/5.6) = P (Z < −2.0) = 0.0227 .

(d)
P (S196 > 90) ≈ P (Z > (90 − 78.4)/5.6) = P (Z > 2.07) = 0.0192 .

(e)
P (S196 > 100) ≈ P (Z > (100 − 78.4)/5.6) = P (Z > 3.857) = 0.00006 .

42.
(a) This is just the sum of four IID exponential random variables:
4
E[S4 ] = .
λ

(b) The conditioning event says that there are two events up to time 1. Hence,
2
E[S4 |N (1) = 2] = 1 + E[time for 2 more events] = 1 + .
λ

(c) Recall that a Poisson process has independent increments. So,

E[N (4) − N (2)|N (1) = 3] = E[N (4) − N (2)] = E[N (2) − N (0)] = E[N (2)] = 2λ .

60. Let {N (t), t ≥ 0} be the Poisson process with rate λ. Then by assumption N (1) = 2.
(a) The probability that both arrived during the first 20 minutes is

2
 P (N (1/3) = 2|N (1) = 2)
P (N (1/3) = 2, N (1) = 2)
=
P (N (1) = 2)
P (N (1/3) = 2, N (1) − N (1/3) = 0)
=
P (N (1) = 2)
P (N (1/3) = 2)P (N (1) − N (1/3) = 0)
= by independent increments
P (N (1) = 2)
1 ( 13 λ)2 − 2 λ
e− 3 λ 2 e
3
= 2
e−λ λ2
1
= .
9
(b) The probability that at least one arrived during the first 20 minutes is

P (N (1/3) ≥ 1|N (1) = 2)

= 1 − P (N (1/3) = 0|N (1) = 2)
P (N (1/3) = 0, N (1) = 2)
= 1−
P (N (1) = 2)
P (N (1/3) = 0, N (1) − N (1/3) = 2)
= 1−
P (N (1) = 2)
P (N (1/3) = 0)P (N (1) − N (1/3) = 2)
= 1− by independent increments
P (N (1) = 2)
1 2 ( 23 λ)2
e− 3 λ e− 3 λ 2
= 1− 2
e−λ λ2
4 5
= 1− = .
9 9

77. in the 9th edition. (The intended problem, not hard.) Using results for a non-
homogenous Poisson process,

P (n events between t=4 and t=5) = e−(m(5)−m(4)) (m(5) − m(4))n /n! = e−11 (11)n /n!

77. in the 10th edition. (Rather difficult.)

Let Tn be the time between the (n−1)st arrival and the nth arrival and let Sn be the service
time of the nth customer.
(a)
µ
P (N = 1) = P (T2 > S1 ) =
µ+λ

(b)

P (N = 2) = P (N > 1)P (N = 2|N > 1)

= P (T2 < S1 )P (min {U1 , U2 } < T3 )

3
 µ ¶µ ¶
λ 2µ
=
λ+µ 2µ + λ
2µλ
=
(2µ + λ)(µ + λ)

where we apply the lack of memory property to get U1 and U2 be i.i.d. exponential random
variables distributed as S1 in line 2.
(c) Similar to the part (b), one can establish that

λ
P (N > j|N > j − 1) = ,
λ + (j − 1)µ

because there are j exponential variables competing to be the next one. That is, we exploit
the lack of memory property to start over after the (j − 1)th arrival. The remaining time to
an arrival remains exponential with mean 1/λ. There are j − 1 customers in service, each of
whom has an exponential remaining service time. The first one to finish is the minimum of
these, which has a rate equal to the sum of the rates. But we then get

P (N > j) = P (N > 1)P (N > 2|N > 1) · · · P (N > j|N > j − 1)

µ ¶µ ¶ µ ¶
λ λ λ
= ··· .
λ+µ λ + 2µ λ + jµ

Then, one can write

P (N = j) = P (N > j − 1) − P (N > j).
(d) Given that there are j customers in the system, since T1 , ..., Tj are i.i.d. exponential,
every customer has the same chance to finish first.
1
P (the f irst customer departs f irst | N = j) =
j
Therefore,
∞
X 1
P (the f irst to arrive is the f irst to depart) = P (N = j)
j=1
j

(e) Let Z be the time until the first departure. Given that N = j, the time until the
first departure is the sum of j + 1 exponential random variables, with successively decreasing
means. In particular, the conditional mean time until the first departure is
j
X 1
E[Z|N = j] =
k=0
λ + kµ

Recall that the minimum itself is exponential with a rate equal to the sum of the rates. Hence,

∞
X
E(time of f irst departure) = E[Z] = E[Z|N = j]P (N = j),
j=1

where E[Z|N = j] is given above.

4
 78. The number of customers that enter the store on a given day is Poisson with mean
Z 17
m(17) − m(8) = λ(t)dt
8
Z 10 Z 12 Z 14 Z 17
= 4 dt + 8 dt + (t − 4) dt + (−2t + 38) dt
8 10 12 14
= 63

94. Let x be a point on the plane and B(x, r) be the circular region with radius r centered
at x. X is the distance from x to its nearest event.
(a)

P (X > t) = P (no event in B(x,t))

= e−λArea (B(x,t)) = e−λπt .
2

(b)
Z ∞
E[X] = P (X > t)dt
Z0∞
2
= e−λπt dt
0
r Z ∞ t2
1 1 −
2 1
= q e 2λπ dt
λ 0 1
2π 2λπ
r
1 1 1
= × = √ .
λ 2 2 λ