8.

The Poisson Distribution

and the Poisson Process
STAT3007: Introduction to Stochastic Processes

Lecturer: Dr. John Wright

1
Contents
• 8.1 The Poisson Distribution
• P.m.f., mean and variance, Law of Rare Events, superposition,
thinning
• 8.2 The Poisson Process
• Definition, independent increments, Example: undersea cable,
Example: customers, non-homogeneity, Example: first aid
• 8.3 The Poisson Point Process
• Little-o notation, the four postulates, definition, model building, PP
vs PPP

2
8.1 The Poisson Distribution
• The Poisson distribution with parameter has p.m.f.

• If has the Poisson distribution, then

• The Poisson distribution is the only distribution with non-

negative integer support whose mean and variance are the
same
3
8.1 The Poisson Distribution
• Consider a large number of independent Bernoulli trials
where the probability of success on each trial is a small
constant
• Let denote the total number of successes in the trials. Then
follows the binomial distribution
•
• Consider the limiting case in which and in such a way that

4
8.1 The Poisson Distribution
• One form of the • Proof
Law of Rare Events
asserts that the
distribution for
becomes the • Recall the definition of :
Poisson • Hence, taking limits as , we have
distribution •.

5
8.1 The Poisson Distribution
• Can Poisson random • Proof
variables be merged?
• Let and be
independent random
variables having Poisson
distributions with
parameters and . • , by the Binomial Theorem.
• Then the sum has a
Poisson distribution
with parameter

6
8.1 The Poisson Distribution
• Can a Poisson random • Proof
variable be thinned?
• Let be a Poisson random
variable with parameter .
Conditional on , let have a
binomial distribution with
parameters and . •.
• Then the unconditional
distribution of is Poisson with
parameter

7
8.2 The Poisson Process
• A Poisson process of intensity or rate is an integer-valued
stochastic process for which
• For any time points , the process increments , are independent
random variables
• For and , the r.v. has the Poisson distribution, i.e. ,

8
8.2 The Poisson Process
• Example
• Defects along an undersea cable occur according to a Poisson
process of rate defects per mile
• What is the probability that no defects appear in the first two
miles of cable?
• Given that there are no defects in the first two miles of cable, what
is the conditional probability of no defect between points two and
three?

9
8.2 The Poisson Process
• What is the probability that no defects appear in the first two miles of
cable?
• Note . Hence
• Given that there are no defects in the first two miles of cable, what is
the conditional probability of no defect between points two and
three?
• Note is independent of . Thus the conditional probability is the same as the
unconditional probability

10
8.2 The Poisson Process
• Example
• Customers arrive in a certain store according to a Poisson process of rate per
hour
• Given that the store opens at 0900, what is the probability that exactly one
customer has arrived by 0930 and a total of five has arrived by 1130?
• Our first task is to translate the question into probability statements
• We are asked to find
• Use independence of increments

11
8.2 The Poisson Process
• Non-homogeneous Poisson processes have all the properties of the
ordinary Poisson process except that the second property
(distribution of the increments) is replaced by

• Where is some deterministic function of time

• We can transform any non-homogeneous Poisson process into a
homogeneous one through a deterministic time change
• Therefore we only need to consider homogeneous Poisson processes

12
8.2 The Poisson Process
• Example
• Demands on a first aid facility in a certain location occur according to a
nonhomogeneous Poisson process having the rate , where is measured in
hours.
• Let . What is the probability two demands occur in the first 2 hours and two
more occur in the second 2 hours?

13
8.2 The Poisson Process
• Solution
• Translate the question into a probability statement. We want to find
• By independent increments this equals
• Work out the average rates over the time intervals first

• Hence the desired probability is

14
8.3 The Poisson Point Process
• Let denote the number of events that occur in the interval
• That is, if denote the times of successive events, then is the
number of values for which
• We will make four postulates about and the events.
• But first, we need some new notation to describe process behavior
over very, very, very short time intervals

15
8.3 The Poisson Point Process
• We say function is (“little o aitch”, or simply “little-o”) if

• Examples
• is
• is not
• Is ?
• , so it is
• If and are both , then and are both
• Intuitively, is if goes to zero faster than
• Now, here come the four postulates
16
8.3 The Poisson Point Process
1. The numbers of events happening in disjoint intervals are
independent r.v.s
• are independent
2. The probability distribution of depends only on and not on
3. There is a positive constant such that the probability of at least one
event happening in a small time interval of length is

4. The probability of two or more events occurring in a small time

interval of length is

17
8.3 The Poisson Point Process
• Notice what the postulates imply about the probability of one event and no event
over a small time interval

• A modeller can check if the phenomenon they are studying satisfies these four
postulates
• Postulate 3 says “events are rare”
• Postulate 4 says “simultaneous events can’t happen”
• In combination with Postulates 3 and 4, Postulates 1 and 2 mean “events happen singly and
discretely” and “only finitely many events can happen in finite time”

18
8.3 The Poisson Point Process
• Many physical phenomena meet these seemingly very general criteria
• It is therefore surprising that these four postulates have a very
specific consequence
• For , for
• with The Four Postulates makes a Poisson Point Process with
intensity
• Building a model from “ground up” (i.e. from postulates) is a very
powerful technique
• And one we will use soon – see Chapter 11

19
8.3 The Poisson Point Process
• What’s the difference between a Poisson Point Process with intensity
and a Poisson Process with rate ?
• counts the number of events occurring in an interval
• counts the number of events occurring up to time
• Basically,
• When the phenomenon we’re modelling are
• events along a line (e.g. occurring on a positive time axis), we can view it as a
Poisson Process or a Poisson Point Process
• events occurring in space (e.g. location of stars in a galaxy), best to view it as
Poisson Point Process

20