STAT3603 Stochastic Processes

Tutorial 9: Poisson Process

Department of Statistics and Actuarial Science
The University of Hong Kong

November 22, 2024

Liu Ruihan [email protected] STAT3603

Review

Interarrival time: The interarrival times {Tn ; n = 1, 2, . . . } in

a Poisson process with rate λ is a sequence of i.i.d.
Exponential(λ) random variables.
Pn
Arrival time or waiting time: Let Sn = i=1 Ti be the
arrival time or waiting time of observing the n-th event. Then
Sn follows a gamma distribution with parameters n and λ
(denoted as Gamma(n, λ)), whose p.d.f. is given by

(λt)n−1
fSn (t) = λe−λt , t ≥ 0.
(n − 1)!

Liu Ruihan [email protected] STAT3603

Review
Conditional distribution of waiting time:
i) Given that N (t) = n, the conditional distribution of the
arrival times (S1 , S2 , · · · , Sn ) is the same as that of
(U(1) , U(2) , · · · , U(n) ), where (U(1) , U(2) , · · · , U(n) ) are the
ordered statistics of n i.i.d. sample of U (0, t). i.e. the
density of S1 , S2 , · · · , Sn |N (t) = n is given by

n!
f (s1 , s2 , . . . , sn |N (t) = n) = 1{0 < s1 < · · · < sn < t}.
tn
ii) Given that Sn = t, the n − 1 arrival times S1 , S2 , · · · , Sn−1
have the same distribution as the order statistics of n − 1
i.i.d. sample of U (0, t).

Liu Ruihan [email protected] STAT3603

Review

Decomposition: Let {N (t); t ≥ 0} be a Poisson process with

rate λ. Suppose that each event can be independently classified
as type j with probability pj , for j = 1, 2, · · · , m. Let Nj (t) be
the number of type j events in [0, t]. Then
{Nj (t); t ≥ 0}, j = 1, 2, · · · , m, are independent Poisson
processes each with rate λj = λpj .

Superposition: Let {Nj (t); t ≥ 0}, j = 1, 2, . . . , m, be m

independent Poisson processes each with corresponding rate λj .
Let N (t) = m
P
j=1 Nj (t), then {N (t); t ≥ 0} is a Poisson process
with rate λ = m
P
j=1 λj .

Liu Ruihan [email protected] STAT3603

Review
Compound Poisson process: A stochastic process
{X(t); t ≥ 0} is said to be a compound Poisson process if it can
be represented as
N (t)
X
X(t) = Yi , t ≥ 0,
i=1

where {N (t); t ≥ 0} is a Poisson process, and {Yi ; i ≥ 1} is a

family of i.i.d. random variables which is also independent of
{N (t); t ≥ 0}. Then we have

E[X(t)] = λtE[Y1 ] and V ar[X(t)] = λtE(Y12 )

Liu Ruihan [email protected] STAT3603

Exercise 1
Suppose that customers arrive at a store in accordance with a Poisson
process having rate 8 per hour; and suppose that each arrival is male
1
with probability 4 and female with probability 34 . Let N1 (t) and
N2 (t) denote respectively the numbers of male arrivals and female
arrivals occurring in [0, t].
(a) What processes do N1 (t) and N2 (t) follow?
(b) It is known that by time t, 5 customers have arrived at the store.
What is the probability that 4 of them are female? And what is
the expected number of female customers among them?
(c) Given that 5 females arrive in the first half hour of opening,
what is the probability that no males arrive in [0, 21 ]? And what
is the expected number of customers who arrive in the first one
hour after opening?
Liu Ruihan [email protected] STAT3603
Solution of Exercise 1
Let N (t) = N1 (t) + N2 (t) be the total number customers by
time t.
(a) N1 (t) and N2 (t) are independent Poisson processes with
rates 8 × 14 = 2 and 8 × 34 = 6 respectively.
(b) The required probability is given by:
P(N2 (t) = 4, N (t) = 5)
P(N2 (t) = 4|N (t) = 5) =
P(N (t) = 5)
P(N2 (t) = 4, N1 (t) = 1)
=
P(N (t) = 5)
P(N2 (t) = 4)P (N1 (t) = 1)
= by independence
P(N (t) = 5)
(6t)4 e−6t 5!
= · 2te−2t ·
4! (8t)5 e−8t
= 0.3955.
Liu Ruihan [email protected] STAT3603
Solution of Exercise 1
(b) The expected number of female customers is naturally
5 × 34 = 3.75. Mathematically, we can show that the
conditional distribution of the number of female customers
is a binomial random variable with parameters 5 and 34 .
P(N2 (t) = n, N1 (t) = 5 − n)
P(N2 (t) = n|N (t) = 5) =
P(N (t) = 5)
(6t)n e−6t (2t)5−n e−2t 5!
= · ·
n! (5 − n)! (8t)5 e−8t
  n 5−n
5 6 2
=
n 85
  n 5−n
5 6 ·2
=
n 8n · 85−n
 
5
= 0.75n 0.255−n , n = 0, 1, 2, 3, 4, 5.
n
Liu Ruihan [email protected] STAT3603
Solution of Exercise 1

(c) Note that N1 (t) and N2 (t) are independent,

P(N1 (0.5) = 0|N2 (0.5) = 5) = P (N1 (0.5) = 0) = e−1 = 0.3679.

Thus, the expected number of customers is

5 + E(N1 (0.5)) + E(N (1) − N (0.5))
= 5 + 2 × 0.5 + 8 × 0.5 = 10

Liu Ruihan [email protected] STAT3603

Exercise 2

Buses arrive at a stop according to a Poisson process (Nt )t≥0

with rate λ. The arrival times are denoted by (Sn )n≥1 . Every
morning you arrive at the bus stop at a fixed time t > 0.
Therefore, SNt is the time when the last bus passed before your
arrival at the stop, and SNt +1 will be the time when you board
in the next bus.

You have a neighbour who is a bit magical: each day he

manages to catch the bus just before the yours.

Liu Ruihan [email protected] STAT3603

Exercise 2

(a) Argue that Vt = SNt +1 − t, your waiting time for your bus,
follows an exponential distribution with rate parameter λ.
(b) (i) By conditioning on Nt , find the cumulative distribution
function of SNt , the departure time of your neighbour.
(ii) Show that E(SNt ) = t − 1
λ + λ1 e−λt .
(c) Calculate E(SNt +1 − SNt ), the average time your neighbour
departs ahead of you.

Liu Ruihan [email protected] STAT3603

Solution of Exercise 2

(a) Recall that Sn := inf{t > 0 : Nt = n}. We have

SNt +1 = inf{s > 0 : Ns = Nt + 1} = inf{s > t : Ns − Nt = 1}

= t + inf{u > 0 : Nt+u − Nt = 1} = t + T̃1 ,

where T̃1 is the first arrival time for a new Poisson process
{Nt+u − Nt , u ≥ 0} with rate λ. Hence, T̃1 = SNt +1 − t
follows exponential distribution with rate parameter λ.

Liu Ruihan [email protected] STAT3603

Solution of Exercise 2
(b) (i) By conditioning on Nt , we find
∞
X
P(SNt ≤ s) = P(SNt ≤ s|Nt = n) · P(Nt = n)
n=0
X∞
= P(Sn ≤ s|Nt = n) · P(Nt = n).
n=0

d
Note that S1 , . . . , Sn |Nt = n ∼ U(1) , . . . , U(n) from U(0, t).
It then follows that
 s n
P(SNt ≤ s|Nt = n) = P(U(n) ≤ s) = .
t
It then follows that
∞  n ∞
X s (λt)n −λt X (λs)n
P(SNt ≤ s) = · e = e−λt · = e−λ(t−s) .
n=0
t n! n=0
n!

Liu Ruihan [email protected] STAT3603

Solution of Exercise 2

(b) (ii) The tail distribution of SNt is given by


1 − e−λ(t−s) , s ≤ t;
P(SNt > s) =
0, s > t.

Hence,
Z ∞
E[SNt ] = P(SNt > s)ds
0
Z t
= 1 − e−λ(t−s) ds
0
1 1
=t− + e−λt .
λ λ

Liu Ruihan [email protected] STAT3603

Solution of Exercise 2

(c) From discussions in (a) and (b), we have

E(SNt +1 − SNt ) = E(SNt +1 − t) + E(t − SNt )

1 1 1
= + t − t + − e−λt
λ λ λ
2 1 −λt
= − e .
λ λ
That is about twice the average elapsed time between two
consecutive buses!

Liu Ruihan [email protected] STAT3603

 – End of Tutorial 9 –

Liu Ruihan [email protected] STAT3603