Chapter 2: Random Variables and Probability Distributions

Since all serious probabilistic computations are performed in terms of random variable,
therefore, let us learn first the concepts of random variables. Alongside, the knowledge of
theoretical probability distribution is of great use in business management and computer
science. The theoretical distribution is divided into two broad categories. Such as discrete and
continuous probability distribution. In this chapter, we will discuss various types of distributions.

1.1 Random Variables

A random variable is a rule that assigns a numerical value to each possible outcome of an
experiment.

Definition:
A random variable ‘X’ on a sample space ‘S’ is a function X:S→R that assigns a real number X(s)
to each sample point s Ꞓ S.

Bernoulli Trails:
Consider a random experiment that has two possible outcomes, “success” and “failure”. Let the
probabilities of the two outcomes be ‘p’ and ‘q’ respectively, with p + q = 1.
Now consider the compound experiment consisting of a sequence of ‘n’ independent repetitions
of this experiment. Such a sequence is known as a sequence of Bernoulli trails.

Example:
Consider a random experiment defined by a sequence of three Bernoulli trails.
S = {000,001,010,100,110,101,011,111}
Now we may define any number of random variables on this sample space.
For example, define a random variable ‘X’ to be the “total number of successes from three trails”.
X(s) = {0,1,1,1,2,2,2,3}
= {0,1,2,3}
For Example, X (0,0,0) = 0 X (1,1,1) = 3
X (1,0,0) = 1 X (1,1,0) =2
From this, we can derive the conclusions that random variable partitions its sample space into a
mutually exclusive and collectively exhaustive set of events.
Thus, for a random variable X and a real number ‘x’, we define the event Ax to be the subset of S
consisting of all sample points ‘S’ to which the random variable X assigns the value of ‘x’:
𝐴 = {𝑠 ∈ 𝑆|𝑋(𝑠) = 𝑥}
Example: 𝐴 = {(0,0,0)}
𝐴 = {(1,0,0), (0,1,0), (0,0,1)}
𝐴 = {(1,1,0), (0,1,1), (1,0,1)}
𝐴 = {(1,1,1)}
Note: 𝐴 ⋂𝐴 = ∅ 𝑖𝑓 𝑥 ≠ 𝑦 𝑎𝑛𝑑 𝑡ℎ𝑎𝑡

𝐴 =𝑆
∈

Types of Random Variables:

There are two types of random variables such as:
1. Discrete Random Variables
2. Continuous Random Variables

The random variables discussed in our example could take an values from a discrete set of
members, hence the image of the random variables is either finite or countable. Such random
variables known as discrete random variables.

Note: A random variable defined on a discrete sample space will be discrete but it is possible to
define a discrete random variable on a continuous sample space.

Exercise:
1. Divide the sample space obtained from an experiment of tossing a coin three mes by
defining all possible random variables.
2. Divide the sample space obtained from an experiment of tossing a die twice by defining
all possible random variables.
1.2 Probability Distribu ons
In this section, we will learn probability mass function and probability density function in
connection to discrete and continuous random variables. Then we will discuss the probability
distributions/cumulative distribution functions.

The Probability Mass Function

The probability mass function (or pmf) or the discrete density function of the random variable X
by
𝑃 (𝑥) = 𝑃(𝑋 = 𝑥)

= 𝑃(𝑠)
( )

= 𝑃𝑟𝑜𝑏𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑡ℎ𝑎𝑡 𝑡ℎ𝑒 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑡ℎ𝑒

𝑟𝑎𝑛𝑑𝑜𝑚 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒 𝑥 𝑜𝑏𝑡𝑎𝑖𝑛𝑒𝑑 𝑜𝑛 𝑎
𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑎𝑛𝑐𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑒𝑥𝑝𝑒𝑟𝑖𝑚𝑒𝑛𝑡 𝑖𝑠 𝑒𝑞𝑢𝑎𝑙 𝑡𝑜 𝑥.
This formula provides us with a method of computing 𝑃(𝑋 = 𝑥)∀ 𝑥 ∈ 𝑅.
Thus we have defined a function with its domain consisting of the event space of the r. v. X, and
with its range in the closed interval [0,1].
The following properties held for the pmf:
(i) 0 ≤ 𝑃(𝑥) ≤ 1 ∀ 𝑥 ∈ 𝑅. This must be true since 𝑃 (𝑥) is a probability.
(ii) Since the random variable assigns some value 𝑥 ∈ 𝑅 to each sample point 𝑠 ∈ 𝑆 , we
must have:
𝑃 (𝑥) = 1
∈

(iii) For a discrete random variable X, the set {𝑥|𝑃 (𝑥) ≠ 0} is a finite or countably infinite
subset of real numbers. Let this set be denoted by {𝑥 , 𝑥 , . . . . . . . . } . the property(ii) be
resorted as
𝑃 (𝑥 ) = 1

Note: A real-valued function 𝑃 (𝑥) defined on R is the pm of some random variable X provided
that it satisfies properties (I) to (III).
Let us verify that 𝑃 (𝑥) hold for X of the above example.
𝑃 (0) = 0.125
 𝑃 (1) = 0.375
𝑃 (2) = 0.375
𝑃 (3) = 0.125
 𝑃 (𝑥)∀𝑥 𝑠𝑎𝑡𝑖𝑠𝑓𝑖𝑒𝑠 𝑝𝑟𝑜𝑝𝑒𝑟𝑡𝑦 (𝑖)
𝑃 (𝑥) = 0.125 + 0.375 + 0.375 + 0.125 = 1
∈

This satisfies property(ii).

Distribution Function:
So far we have restricted our attention to computing P(X=x), but often we may be interested in
computing the probability of the set {𝑠|𝑋(𝑠) ∈ 𝐴} = ⋃ {𝑠|𝑋(𝑠) = 𝑥 } (1)
This event is denoted as [𝑋 ∈ 𝐴] and its probability by 𝑃(𝑋 ∈ 𝐴).
Let A=(a, b), then we write 𝑃(𝑎 < 𝑋 < 𝑏).
Similarly, if 𝐴 = (𝑎,𝑏] then 𝑃(𝑋 ∈ 𝐴) will be written as 𝑃(𝑎 < 𝑋 ≤ 𝑏).
If 𝑃 (𝑥) denotes the pmf of r.v. X, then from equation(1) we have:

𝑃(𝑋 ∈ 𝐴) = 𝑃 (𝑥 )
∈

The semi-infinite interval 𝐴 = (−∞, 𝑥) will be of special interest and in this case we denote as
[𝑋 ≤ 𝑥].
For example: 𝑃(𝑋 ≤ 2) = 𝑃(𝑥 = 0) + 𝑃(𝑥 = 1) + 𝑃(𝑥 = 2)
= 𝑃 (0) + 𝑃 (1) + 𝑃 (2)
1 3 3 7
= + + =
8 8 8 8
The function Fx(t),-tdefined by
Fx(t)= P(-< x t)
= P(x t)
= Px(x)
xt

is called the probability distribution function or the cumulative distribution function (CDF) of the
random variable X.

P(a < x b)= P(x b)-P(x a) =F(b)-F(a)

Some of the properties of F(x):

1. 0Fx(x) fon -<x<This follow because F(x) is a probability.

2. If x1x2,then F(x1)F(x2)
3. lim F(x)=1, and lim F(x)=0
xx
4. F(x) has a positive jump equal to Px(xi) at i=1,2,....... and in the interval(xi,xi+1), F(x) has a
constant value. Thus
F(x)=F(xi) for xi x <xi+1
and F(xi+1)=F(xi)+Px(xi+1)

Note:
It can be shown that any function F(x) satisfying properties (i) to (iv) is the distribution function
of some discrete r.v.

Let us summarize the domain and range of P, X, pmf, and CDF through the below given figure.

X
S R

Px(x) F(x)

ƒ [0,1]

[Link] Representation to Understand r.v., pmf, PDF.

Exercise:

1. Define a random variable on the sample space obtained from an experiment of tossing a
coin 5 times. Find the probability mass function on X.
Some Popular Discrete Distributions:
Let us learn some of the popular probability mass functions and rspective
distribution functions.
i. The Bernoulli pmf and CDF
ii. The Binomial pmf and CDF
iii. The Poisson pmf and CDF

The Bernoulli pmf/CDF:

The Bernoulli pmf is the density function of a discrete r.v. X having ‘0’ and ‘1’ as
its only possible value; it originates from the experiment consisting of a single
Bernoulli trial. It is given by
Px(0)= P0= P(X=0)= q
Px(1)= P1= P(X=1)= p, where p+q=1

The corresponding CDF is given by

0 for x<0
Fx(x) = q for 0x<1
1 for x1

The Binomial pmf/CDF:

To generate the Bernoulli pmf, we considered a single Bernoulli trial. Now

consider a sequence of “n” independent Bernoulli trial with the probability of
success equal to p on each trial.

Let Yn be the r.v. defined as the number of successes in “n” trials. The domain of
the r.v. Yn is all the n tuples of “0” and “1” and the images is 0,1,2,......,n

he pmf of Yn is:
nck. pk(1-p)n-k for 0n
PYn(k)=
0 otherwise

This gives us the probability of k ‘success’ in ‘n’ independent trials of an

experiment that has probability ‘p’ of an experiment.
The Binomial density with parameters ‘n’ and ‘p’ often denoted by b(k; n, p).

The pmf PYn(k) is holding the properties of pmf.

The property one can be easily verified. Let us check whether it is satisfying
property-2 or not.
n
i.e.,  PYn(k)= 1
k=0

n
LHS=  Pyn(k)
k=0

n
=  nck . pk (1-p)n-k
k=0

= p+(1-p)n (∵by Binomial theorem)

= 1n
= 1
= R.H.S.

This is the reason for the name Binomial pmf. The cumulative distribution
function of a Binomial random variable will be denoted by B(t; n, p) and is given
by
B(t, n, p)= FYn(t)
t
=  (nci).pi(1-p)n-i
i=0

Note:
The Binomial distribution is applicable whenever a series of trials is made
satisfying the following conditions
1. Each trial has exactly two mutually exclusive outcomes, usually labeled
“success” and “failure”.
2. The probability of “SUCCESS” on each trial is a constant, denoted by ‘p’.
The
probability of failure is q=1-p.
3. The outcomes of successive trials are mutually independent.
Important Note:
The binomial model has two parameters, n and p. If the values of model
parameters are known, then it is relatively easy to evaluate the probabilities of
the events of interest.

Exercise:
1. A coin is tossed SIX times, what is the probability of obtaining four or more
heads?
Solution: When a coin is tossed the probabilities of head and tail in case of an
unbiased coin are equal i.e. p = q = .

The probability of obtaining 4 heads:

= nck . [Link]-k = 6c4 ( )4 . ( )2

= 15 . . = (∵ 𝑛 = 6, 𝑘 =
4)
= 0.234
The probability of obtaining 5 heads:

= 6c5( )5 . ( ) = 6 = 0.094

The probability of obtaining 6 heads:

= 6c6( )6 = 1. = = 0.01562

Hence, the probability of getting 4 or more heads:

= 0.234 + 0.094 + 0.01562 = 0.34362
2. The incidence of COVID-19 disease is such that on the average 20% of West
Bengal returnees with no symptoms of COVID-19 suffer from it. If 10 West
Bengal returnees with no symptoms of COVID-19 are selected at random, find
the probability that
i. Exactly 2 W.B. returnees with no symptoms of COVID-19 suffer from
COVID-19.
ii. Not more than 2 W.B. returnees with no symptoms suffer from COVID-
19.

Solution: Probability that a W.B. returnees with no symptoms of COVID-19.

⇒ = = =𝑃

⇒ q = 1- =

According to binomial model, the probability that out of 10 W.B. returnees, x

returnees with no symptoms of COVID-19 suffer from COVID-19 = 100cx . ( )x.
( )10-x

i. X=2 10c2. ( )2 . ( )10-2 = 45. ( ) . ( )8 =0.302

ii. The required probability that no more than 2 W.B. returnees with no
symptoms of COVID-19 is given by:
=p (0) + p (1) + p (2)
 =10c0. ( )0 ( )10 +10c1. ( ). ( )9 + 10c2 ( )2. ( )8

= { 410+ 10.49 + 45.48}

( )

= {410 + 10.49 + 45.48} = 0.678

( )

Note: When ‘n’ becomes very large, computation using the binomial formula
becomes unmanageable. In the limit as n → ∞ , it can be shown that:
( )
b (k; n, p) = .𝑒

This is known as Laplace (or normal) approximation to the Binomial pmf and the
agreement between two formula’s depends on the values of ‘n’ and p.

It is suggested that for n≥10, if ≤𝑃≤ then the Laplace formula

provides a good approximation to the Binomial pmf.

The Poisson pmf/CDF:

This distribution was developed by a French mathematician, Simeon Denis
Poisson in 1837. Poisson distribution may be expected in cases where the
chance of any individual event being a success is small. The distribution is used
to describe the behaviour of rare events such as the member of accidents on
road, no. Of printing mistakes in a book, etc.
Let us consider another problem related to the binomial distribution
Suppose we are observing the arrival of jobs to a large computation center for
the time interval (0, 𝑡]. It is reasonable to assume that for each small interval of
time ∆t the probability of a new job arrival is 𝜆. ∆𝑡 , where ′𝝀′ 𝑖𝑠 a constant that
depends upon the user population of the computation center. If the ∆t is
sufficiently small, then the probability of two or more jobs arriving in the interval
of duration ∆t may be neglected.
We are interested in calculating the probability of 'k’ jobs arriving in the
interval of duration ‘t’.
Suppose that the interval (0,t) is divided into ‘n’ subintervals of length t/n, and
suppose further that the arrival of a job in any given interval independent of the
arrival of a job in any other interval. Then for a sufficiently length n, we can think
of the ‘n’ intervals as consisting a sequence of Bernoulli trials with the probability
of success p = 𝜆t/n.
It follows that the probability of ‘k’ arrivals in a total of ‘n’ intervals each with a
duration t/n is approximately given by :

b(k; n, ) = nck . ( ) . (1- )n-k k = 0,1,2,……….,n

!
= )!
. ( )k . (1- )n-k
!(

. ………..( )( )……..( )( ). ( )
= )!
. . (1 - )n-k
!(

=( )……( )( ). .
(𝜆𝑡) . (1 - )n. (1 - )-k
!.

= . . ……. . ( ( )) . ( )
. (1 - )n. (1 - )-k
!

Take limiting case as n -> ∞

b(k;n, )

( ) ( )
= .( 1 - ) .( 1 - )…….. . (1 - ). . (1 - )n. (1 - )-k
!

( )
= 1 .1 ……. 1. . lim 1 − .1
! →

( )
= . lim 1 −
! →

( )
= . lim { 1 − }
! →

Let h = − , n -> ∞ , h -> 0

( )
= . { lim (1 + ℎ) }
! →

( )
= 𝑒 ( Since the limit in the bracket is the common
!
definition of ‘e’ .)

Thus the Binomial pmf approaches

.( )
b(k;n, )= K =0 , 1 , 2 ,………..
!

now replacing 𝜆𝑡 ′ by a single parameter ‘ ∝ ‘ .

We get the well known poisson pmf:

 ∝ .∝
f(k:∝)= , k = 0 , 1 , 2 ,…….
!

Thus the poisson pmf can be used as a convenient approximation to the binomial
pmf when ‘n’ is large and ‘p’ is small.
∝ .∝
n
ck .pk qn-k = , [where ∝ = np]
!

An acceptable rule of thumb is to use poisson approximation for binomial

probabilities if ‘n’ >= 20 and ‘p’ <= 0.05 .

NOTE: Lets verify whether f(k : ∝) is satisfying the properties of pmf.

The first property of pmf is easily derived that f(k : ∝) are non-negative and the
value for each k is lies between [ 0 ,1 ].
The second property says that the sum is equal to 1 i.e.,
∑ f(k ∶∝)=1
L.H.S=
∝ .∝
∑ f(k ∶∝)=∑
!
∝
𝑒 ∝∑
=𝑒 ∝
. 𝑒 ∝ = 𝑒 =1
!

Exercise:
1. A manufacture purchase IC chips, 1 percent of which are defective. Find the
probability that in a box containing 100 chips, no defectives are found.
Solution :
Here n = 100 , p = 0.01 , according to binomial probability mass
function
100
b ( 0; 100 , 0.01 ) = c0 ( 0.01)0 (0.99)100
= ( 0.99 )100 =0.366
However, according to poisson approximation,
We get f ( k ∶ ∝ ) = f ( 0 ; 100 , 0.01 )
f ( 0 , 1) = e -1 = 0.3679
NOTE : In many situations, the poisson pmf is very useful .

In reliability theory, it is quite reasonable to assume that the probability of k

component malfunctioning within an interval of time ‘t’ in a system with a large
number of component is given by the poisson pmf (here 𝜆 is known as the
component failure rate):
( )
f( k : ∝ ) = 𝑒 , k = 0 ,1 ,2 ,………
!

Probability density function:

For the case of continuous random variable, X, the probability density function
is denoted as 𝑓 (𝑥). It is defined as 𝑓 (𝑥) = 𝐹 (𝑥), where x is between −∞
to +∞.

The pdf, 𝑓 (𝑥), satisfies the following properties.

Distribution Function:
The distribution function 𝐹 (𝑥) on the continuous random variable is defined
as:

𝐹 (𝑥) = 𝑃(𝑋 ≤ 𝑡) = ∫ 𝑓 (𝑥)𝑑𝑥 , where t is between −∞ to +∞.

The other probabilities of interest can also be obtained as:

Like the CDF of discrete random variable, the CDF of continuous random
variable satisfies the following properties.
Example:
The time (measured in years), X , required to complete a software project has a
pdf of the form:

What is the probability that the project will be completed in less than 4
months?
Solution:

⇒ 𝑓 (𝑥)𝑑𝑥

+ 𝑓 (𝑥)𝑑𝑥

+ 𝑓 (𝑥)𝑑𝑥 = 1

⇒0+ 𝑘𝑥(1 − 𝑥)𝑑𝑥 + 0

=1

⇒𝑘 𝑥(1 − 𝑥)𝑑𝑥 = 1

⇒𝑘 (𝑥 − 𝑥 )𝑑𝑥 = 1

⇒ 𝑘 ∫ 𝑥𝑑𝑥 − ∫ 𝑥 𝑑𝑥 = 1

⇒𝑘 − =1

⇒𝑘 − =1

⇒ =1

⇒𝑘=6
Now we can find the probability that the project will be completed in less than
four months is given by:

𝑃 𝑋< = 𝑃(𝑋 < )

 =𝐹 = ∫ 𝑓 (𝑥)𝑑𝑥

= ∫ (6𝑥 − 6𝑥 )𝑑𝑥 = 6 −6

=3 −0 −2

= = or about a 26% chance.

Exercise:
1. Find the value of the constant k so that:
𝑘𝑥 (1 − 𝑥 ), 0 < 𝑥 < 1,
𝑓(𝑥) =
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒,
is a proper density function of a continuous random variable.

Some of the Special Distributions:

1. The Exponen al Distribu on:
This distribution, sometimes called the negative exponential distribution, occurs
in applications such as reliability theory and queuing theory. Reasons for its use
include its memoryless (Markov) property and resulting analytical tractability
and its relation to the (discrete) Poisson distribution.
Thus the following random variables will often be modelled as exponential:
1) Time between two successive job arrivals to a computing center (often
called interarrival time)
2) Service time at a server in a queuing network; the serves could be a
resource such as the CPU, I/O device, or a communication channel.
3) Time to failure (lifetime) of a component.
4) Time required to repair a component that has malfunctioned.
The exponential distribution function, shown in Figure 2 is given by:
 Fig.2. Exponential Distribution Function

1−𝑒 𝑖𝑓 0 ≤ 𝑥 ≤ ∞ (2)
𝐹 (𝑥) =
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
If a random variable X possesses CDF given by equation (2), we use the notation
X˜̃EXP(λ) for brevity. The Pdf of X has the shape shown in fig.(below) and is given
by

𝜆𝑒 𝑖𝑓 𝑥 > 0
𝑓(𝑥) = (3)
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

(∵ 𝑓(𝑥) = 𝐹 (𝑥) )

The exponential pdf for different value of λ is as follows.

Let us verify that this pdf is satisfying the following properties.

It is obvious from (3) that
𝑓(𝑥) ≥ 0 ∀𝑥
We now verify for 2nd property

∫ 𝜆𝑒 𝑑𝑥 = 1
L.H.S = ∫ 𝜆𝑒 𝑑𝑥 = 𝜆 ∫ 𝑥 𝑒 𝑑𝑥
( )
=𝜆 (According to gamma function)

= Γ(1) (∵ Γ(1) = (1 − 1)! = 0! = 1

= 1 =R.H.S

Normal or Gaussian Distribution:

The normal distribution, also called the normal probability happens to be most
useful theoretical distribution for continuous variables. Many statistical data
concerning business and economic problems are displayed in the form of normal
distribution.
Normal distribution was discovered by Gauss in1809 and by Laplace in [Link]
normal model has, nevertheless, become the most important probability model
in statistical analysis.
The normal distribution is an approximation to binomial distribution. The normal
distribution is extremely important in statistical applications because of the
central limit theorem, which state that, under very general assumptions, the
mean of the sample of ‘n’ mutually independent random variables (having
distribution with finite mean and variance) is normally distributed in the limit
n→ ∞.
It has been observed that errors of measurement often possess this distribution.
The normal density has the well known bell shaped curve and is given by;
( )
𝑓 (𝑥) = .𝑒 ,-∞ < 𝑥 < ∞,
√

where −∞ < 𝑥 < ∞ and 𝜎 > 0 are two parameters of the distribution. If a
random variable x has the pdf (as above), then we write x of (𝜇, 𝜎 ).
The standard normal distribution [Z~𝑁(0,1)] a normal distribution with
parameters 𝜇 = 0 and 𝜎 = 1 is defined as
( )
𝐹 = ∫ 𝑒 dt
√

Let us visualize the normal the normal density with parameters 𝜇 = 2 and 𝜎 =
1.
 Fig.3. Normal density with parameters μ = 2 and σ = 1
Note:
(1) The normal distribution can have different shapes depending on different
values of 𝜇 and 𝜎 but there is one and only one normal distribution for any
given pair of values for 𝜇 and 𝜎.
(2) Normal distribution is a limiting case of binomial distribution when n→ ∞
and neither p nor q is very small.
(3) Normal distribution is limiting case of Poisson distribution when its mean is
large.
(4) The mean of a normally distributed population lies at the center of its
normal curve.
(5) The two tails of normal pdf extend indefinitely and never touch the
horizontal axis.
(6) Since there is only one maximum point, the normal curve is unimodal, i.e. it
has only one mode.

Uniform Distribution:
The random variable 𝑥 is said to have a continuous uniform distribution if its
probability density function 𝑓(𝑥) is constant i.e., k over the interval (a,b) and
zero, elsewhere. Therefore,

𝑘; 𝑎<𝑥<𝑏
𝑓(𝑥) =
0 ; 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

Since the area under the density curve, i.e., total

Probability is one, we have

∫ 𝑓(𝑥) 𝑑𝑥 = 1
⇒ ∫ 𝑘 𝑑𝑥 = 1
⇒ k(𝑏 − 𝑎) = 1
⇒ k=( )
Hence, the probability density function of uniform distribution is given as

1
𝑖𝑓 𝑎 < 𝑥 < 𝑏
𝑓(𝑥) = (𝑏 − 𝑎)
0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

The uniform distribution is otherwise known as rectangular distribution.