Stochastic Processes ( 2) \ Statistics & Informatics dep. \Fourth class.

(2020-2021)

Assistant prof. Dr. Muthanna Subhi Sulaiman \University of Mosul.

Stochastic Processes (2)

Lecture 5: Poisson process:

(5.1) Arrival Process:

Let 𝑡 represent a time variable. Suppose an experiment

begins at 𝑡 = 0. Events of a particular kind occur randomly,
the first at 𝑡1, the second at 𝑡2 , and so on. The random variables
𝑡𝑖 denotes the time at which the 𝑖 𝑡ℎ event occurs, and the values
𝑡𝑖 (𝑖 = 1,2, . . . ) are called points of occurrence.

Then the process {𝑇𝑛 , 𝑛 ≥ 0} denotes the time from the

beginning until the occurrence of the 𝑛𝑡ℎ event is called an
arrival process.

(5.2) Interarrival Process

Let: 𝑋𝑛 = 𝑡𝑛 − 𝑡𝑛−1 and 𝑡0 = 0. Then 𝑋𝑛 denotes the

time between the (𝑛 − 1)𝑡ℎ and the 𝑛𝑡ℎ events. The sequence
of ordered r.v.'s {𝑋𝑛 , 𝑛 ≥ 1} called an interarrival process. If
all random variables 𝑋𝑛 are independent and identically
distributed, then {𝑋𝑛 , 𝑛 ≥ 1} is called a renewal process or a
recurrent process.

𝑋1 𝑋2 𝑋3 𝑋𝑛

Where: 𝑇𝑛 = 𝑋1 + 𝑋2 + ⋯ + 𝑋𝑛

1
Stochastic Processes ( 2) \ Statistics & Informatics dep. \Fourth class. (2020-2021)

Assistant prof. Dr. Muthanna Subhi Sulaiman \University of Mosul.

(5.3) Counting Process:

A stochastic process {𝑁𝑡 : 𝑡 ≥ 0} is a counting process if
𝑁𝑡 represents the total number of “events” that have occurred
in the interval (0, 𝑡). Hence counting process {𝑁𝑡 } should
satisfy:
1. 𝑁𝑡 ≥ 0 and 𝑋0 = 0
2. 𝑁𝑡 is integer valued
3. If 𝑠 < 𝑡, then 𝑁𝑠 ≤ 𝑁𝑡
4. For 𝑠 < 𝑡, 𝑁𝑡 − 𝑁𝑠 equals the number of events that have
occurred on the interval (𝑠, 𝑡].
Examples for counting process:
• Number of persons entering a store before time 𝑡
• Number of people who were born by time 𝑡
• Number of goals a soccer player scores by time 𝑡.
• The location of users in a wireless network
A counting process 𝑁𝑡 is said to possess independent
increments if the numbers of events which occur in disjoint
time intervals are independent.
A counting process 𝑁𝑡 is said to possess stationary increments
if the distribution of the number of events which occur in any
interval of time depends only on the length of the time interval.
i.e. the number of events in the interval
(𝑠 + ℎ, 𝑡 + ℎ) that is, 𝑁𝑡+ℎ − 𝑁𝑠+ℎ has the same distribution
as the number of events in the interval (𝑠, 𝑡) that is, 𝑁𝑡 − 𝑁𝑠
for all 𝑠 < 𝑡 and ℎ > 0.

2
Stochastic Processes ( 2) \ Statistics & Informatics dep. \Fourth class. (2020-2021)

Assistant prof. Dr. Muthanna Subhi Sulaiman \University of Mosul.

(5.4) Poisson process:

One of the most important types of counting processes is the
Poisson process (or Poisson counting process), which is
defined as follows:
Definition:
The counting process {𝑁𝑡 : 𝑡 ≥ 0} is said to be a Poisson process
having rate (intensity) 𝜆, (𝜆 > 0) if:
1. 𝑁0 = 0 .
2. The process has independent increments.
3. The number of events in any interval of length 𝑡 is Poisson
distributed with mean 𝜆𝑡 ; that is, for all 𝑠, 𝑡 > 0,

It follows from condition (3) that a Poisson process has

stationary increments and that:

𝐸[𝑁𝑡 ] = 𝜆𝑡

and 𝑉𝑎𝑟[𝑁𝑡 ] = 𝜆𝑡

3
Stochastic Processes ( 2) \ Statistics & Informatics dep. \Fourth class. (2020-2021)

Assistant prof. Dr. Muthanna Subhi Sulaiman \University of Mosul.

Thus, the expected number of events in the unit interval (0, 𝑡),
or any other interval of unit length, is just 𝜆 .

(5.5) The Second Definition of a Poisson Process:

The counting process {𝑁𝑡 : 𝑡 ≥ 0} is said to be a Poisson

process having rate (intensity) 𝜆 , (𝜆 > 0) if:

1. 𝑁0 = 0 .
2. The process has independent and stationary increments.
3. We have:

▪ 𝑃{𝑁𝑡+Δ𝑡 − 𝑁𝑡 = 0} = 𝑃{𝑁𝑡,𝑡+𝛥𝑡 = 0} = 1 − 𝜆 Δ𝑡 + 𝑂(Δ𝑡)

▪ 𝑃{𝑁𝑡+Δ𝑡 − 𝑁𝑡 = 1} = 𝑃{𝑁𝑡,𝑡+Δ𝑡 = 1} = 𝜆 Δ𝑡 + 𝑂(Δ𝑡)

▪ 𝑃{𝑁𝑡+Δ𝑡 − 𝑁𝑡 ≥ 2} = 𝑃{𝑁𝑡,𝑡+Δ𝑡 ≥ 2} = 𝑂(Δ𝑡)

where 𝑂(Δ𝑡) is a function of Δ𝑡 which goes to zero faster than

𝑂(Δ𝑡)
does Δ𝑡; that is: lim =0
∆𝑡→0 Δ𝑡

Proof of point 3:
▪ The probability that no event occurs in any short interval
approaches unity as the duration of the interval approaches
zero.
𝑃{𝑁𝑡+Δ𝑡 − 𝑁𝑡 = 0} = 𝑒 −𝜆Δ𝑡
𝜆2 2
= 1 − 𝜆 Δ𝑡 + Δ𝑡 − ⋯, (Taylor Series)
2!

= 1 − 𝜆 Δ𝑡 + 𝑂(Δ𝑡)

4
Stochastic Processes ( 2) \ Statistics & Informatics dep. \Fourth class. (2020-2021)

Assistant prof. Dr. Muthanna Subhi Sulaiman \University of Mosul.

Note that if Δ𝑡 is small, the terms that include second or

higher powers of Δ𝑡 are negligible compared to Δ𝑡.
▪ Now, let us look at the probability of having one arrival in
an interval of length Δ𝑡.
𝑃{𝑁𝑡+Δ𝑡 − 𝑁𝑡 = 1} = 𝜆 Δ𝑡 𝑒 −𝜆Δ𝑡
𝜆2 2
= 𝜆 Δ𝑡(1 − 𝜆 Δ𝑡 + Δ𝑡 − ⋯ ), (𝑇𝑎𝑦𝑙𝑜𝑟 𝑆𝑒𝑟𝑖𝑒𝑠)
2!

𝜆3
= 𝜆 Δ𝑡 + (−𝜆2 Δ𝑡 + 2
Δ𝑡 3
− ⋯)
2!

= 𝜆 Δ𝑡 + 𝑂(Δ𝑡)
▪ Similarly, we can show that:𝑃{𝑁𝑡+Δ𝑡 − 𝑁𝑡 ≥ 2} = 𝑂(Δ𝑡)
It can be shown that in the Poisson process, the intervals
between successive events are independent and identically
distributed exponential r.v.'s. Thus, we also identify the Poisson
process as a renewal process with exponentially distributed
intervals.

The counting process {𝑁𝑡 : 𝑡 ≥ 0} is called a Poisson process if

the inter-arrival times 𝑋1 , 𝑋2 , … follow the exponential
distribution.
We conclude that the Poisson process is a discrete state,
continuous time process.

5
Stochastic Processes ( 2) \ Statistics & Informatics dep. \Fourth class. (2020-2021)

Assistant prof. Dr. Muthanna Subhi Sulaiman \University of Mosul.

Example (1):
The number of customers arriving at a market can be modeled
by a Poisson process with intensity 𝜆 = 12 customers per hour.
1. Find the probability that there are 2 customers between
10: 00 and 10: 20.
2. Find the probability that there are 3 customers between
10: 00 and 10: 20 and 7 customers between 10: 20 and
11.
Solution:
1. Here, 𝜆 = 12 and the interval between 10: 00 and 10: 20
has length
𝑡 = 20 minutes. Thus, if 𝑋 is the number of arrivals in that
12
interval, we can write 𝑁𝑡 ∼ 𝑃𝑜𝑖𝑠𝑠𝑜𝑛 (60) per minutes.
Therefore:
𝑒 −𝜆𝑡 (𝜆𝑡)𝑛
𝑃𝑟 {𝑁(𝑡) = 𝑛} = , 𝑛 = 0,1,2, …
𝑛!
12 2
− (20) 12
𝑒 60 ( (20))
𝑃𝑟 {𝑁(20) = 2} = 60
= 8𝑒 −4 = 0.147
2!

2. We have two non-overlapping intervals:

𝐼1 = (10: 00 𝑎. 𝑚. , 10: 20 𝑎. 𝑚. ] and
𝐼2 = (10: 20 𝑎. 𝑚. , 11 𝑎. 𝑚. ].
Thus, we can write:

6
Stochastic Processes ( 2) \ Statistics & Informatics dep. \Fourth class. (2020-2021)

Assistant prof. Dr. Muthanna Subhi Sulaiman \University of Mosul.

𝑃(3 𝑎𝑟𝑟𝑖𝑣𝑎𝑙𝑠 𝑖𝑛 𝐼1 & 7 𝑎𝑟𝑟𝑖𝑣𝑎𝑙𝑠 𝑖𝑛 𝐼2 ) =

𝑃(3 𝑎𝑟𝑟𝑖𝑣𝑎𝑙𝑠 𝑖𝑛 𝐼1 )𝑃(7 𝑎𝑟𝑟𝑖𝑣𝑎𝑙𝑠 𝑖𝑛 𝐼2 )
(Independent increment), then:
𝑃𝑟 {𝑁(20) = 3}. 𝑃𝑟 {𝑁(40) = 7}
3 7
12 12
12 ( (20)) 12 ( (40))
− (20) 60 − (40) 60
= 𝑒 60 . 𝑒 60
3! 7!
= 4438.42 𝑒 −4 𝑒 −8
= 4438.42 𝑒 −12
= 0.027

Example (2):

Defects occur along an undersea cable according to a Poisson

process of rate 𝜆 = 0.1 per mile.
1. What is the probability that no defects appear in the first
two miles of cable?
2. Given that there are no defects in the first two miles of
cable, what is the conditional probability of no defects
between mile points two and three?
To answer (1) we observe that 𝑁(2) has a Poisson distribution
whose parameter is (0.1)(2) = 0.2. Thus:
𝑃𝑟 {𝑁(2) = 0} = 𝑒 −0.2 = 0.8187.

7
Stochastic Processes ( 2) \ Statistics & Informatics dep. \Fourth class. (2020-2021)

Assistant prof. Dr. Muthanna Subhi Sulaiman \University of Mosul.

In part (2) we use the independence of 𝑁(3) − 𝑁(2) and

𝑁(2) − 𝑁(0) = 𝑁(2). Thus, the conditional probability is the
same as the unconditional probability, and
𝑃𝑟 {𝑁(3) − 𝑁 (2) = 0} = 𝑃𝑟 {𝑁(3 − 2) = 0}
𝑃𝑟 {𝑁(1) = 0} = 𝑒 −0.1 = 0.9048 (Poisson distribution)

Example (3):

Customers arrive in a certain store according to a Poisson

process of rate 𝜆 = 4 per hour. Given that the store opens at
9: 00 𝐴. 𝑀., what is the probability that exactly one customer
has arrived by 9: 30 and a total of five have arrived by
11: 30 𝐴. 𝑀.?
Measuring time 𝑡 in hours from 9: 00 𝐴. 𝑀. , we are asked to
determine:
1 5
𝑃𝑟 {𝑁 (2) = 1, 𝑁 (2) = 5}
5 1 1
We use the independence of 𝑁 (2) − 𝑁 (2) and 𝑁 (2) to
reformulate the question thus:
1 5 1 5 1
𝑃𝑟 {𝑁 ( ) = 1, 𝑁 ( ) = 5} = 𝑃𝑟 {𝑁 ( ) = 1, 𝑁 ( − ) = 5 − 1}
2 2 2 2 2

1
= 𝑃𝑟 {𝑁 (2) = 1 } . 𝑃𝑟 {𝑁(2) = 4}
1 1
−(4) 1 4
𝑒 2 ((4) )
2 𝑒 −(4)2 ((4)2)
={ }.{ }
1! 4!

= 0.0155.
8
Stochastic Processes ( 2) \ Statistics & Informatics dep. \Fourth class. (2020-2021)

Assistant prof. Dr. Muthanna Subhi Sulaiman \University of Mosul.

Example (4):
Suppose that the process 𝑁(𝑡) has a Poisson process with rate
𝜆 = 8 , then the probability:
𝑃𝑟 {𝑁 (2.5) = 17, 𝑁(3.7) = 22, 𝑁(4.3) = 36}
= 𝑃𝑟 {𝑁(2.5) = 17, 𝑁(3.7 − 2.5) = 22 − 17, 𝑁(4.3 − 3.7) = 36 − 22}

= 𝑃𝑟 {𝑁 (2.5) = 17} . 𝑃𝑟 {𝑁 (1.2) = 5} . 𝑃𝑟 {𝑁 (0.6) = 14}

17 5 14
𝑒 −(8)2.5 ((8)2.5) 𝑒 −(8)1.2 ((8)1.2) 𝑒 −(8)0.6 ((8)0.6)
={ 17!
}.{ 5!
}.{ 14!
} = 0.000001