UNIT 6

BIVARIATE RANDOM
VARIABLE

Structure
6.1 Introduction 6.5 Bivariate Continuous
Random Variable
Expected Learning Outcomes
6.6 Summary
6.2 Bivariate Random Variable
6.7 Terminal Questions
6.3 CDF for Bivariate Random
Variable 6.8 Solutions/Answers
6.4 Bivariate Discrete Random
Variable

6.1 INTRODUCTION
In Units 4 and 5, you have studied discrete and continuous univariate random
variables and their different distributions like CDF, PMF, PDF, etc. In the
present unit, we will discuss bivariate discrete and continuous random
variables and their corresponding distributions. First of all, what we mean by a
bivariate random variable is discussed in Sec. 6.2. Like univariate case CDF is
also defined in the same way in bivariate discrete and continuous worlds of
probability theory and is defined in Sec. 6.3. The next two sections, i.e., Sec.
6.4 and Sec. 6.5 discuss discrete and continuous bivariate random variables
respectively.

What we have discussed in this unit is summarised in Sec. 6.6. Self-

Assessment Questions (SAQs) have been given in some sections which are
generally based on the content discussed in that section. But to give you a
good practice of what we have discussed in this unit some more questions
based on the entire unit are given in Sec. 6.7 under the heading Terminal
Questions. Due to the reason mentioned in Sec. 1.1 of Unit 1 of this course,
solutions of all the SAQs and Terminal Questions are given in Sec. 6.8.

In the next unit, you will study what is expected value of a distribution and
moment generating function (MGF) of the random variable.

Expected Learning Outcomes

After completing this unit, you should be able to: 135
Probability ❖ explain the concept of discrete and continuous bivariate random
Measure and variables;
Random
Variable ❖ define joint, marginal and conditional probability mass/density functions
and CDF in bivariate random variables setting; and

❖ able to apply the concept of PMF, PDF and CDF to obtain required
probabilities in bivariate setting.

6.2 BIVARIATE RANDOM VARIABLE

In Unit 2, you have studied what is a probability space and in Unit 4, you have
studied random variable. Recall those concepts because in this unit, we will
use many concepts related to probability measure and random variable.
Let ( , , ) be a probability space. Let X :  → and Y :  → be two
random variables. Recall the definition of univariate random variable which
associates a unique real number to each member of  . Note that both X and
Y are defined on the same sample space , so the ordered pair
( X(), Y()) , of real numbers for each    lies in 2 =  . So, the
ordered pair random variable (X, Y) which associates a unique ordered pair of
real numbers to each member of the sample space is called bivariate random
variable on  , i.e., (X, Y) is a function which can be denoted as follows.
( X(.), Y(.)) :  → 2

For example, consider the random experiment of tossing a coin twice. Sample
space of this random experiment is  = {HH, HT, TH, TT}. Let (X, Y) be the
bivariate random variable, where X denotes the number of heads in two tosses
and Y denotes the number of tails before the first head. So, the bivariate
random variable (X, Y) associates an ordered pair of real numbers (2, 0) with
HH, (1, 0) with HT, (1, 1) with TH and (0, 2) with TT and is visualised in Fig.
6.1 as follows.

Fig. 6.1: Visualisation of the bivariate random variable (X, Y) as a function from the
2
sample space  to xy-plane  =

In Unit 4, you have studied what is a discrete random variable and in Unit 5,
you have studied what is a continuous random variable. If both X and Y are
discrete random variables then the ordered pair random variable (X, Y) is
called a discrete bivariate random variable. However, if both X and Y are
continuous random variables then it is not necessary that (X, Y) will be joint
136 continuous bivariate random variable. Why it is so its discussion is beyond the
scope of the course. The technical word is jointly continuous which is used Bivariate Random
when we want that bivariate random variable (X, Y) is continuous. So, if both X Variable

and Y are jointly continuous random variables then the ordered pair random
variable (X, Y) is called a continuous bivariate random variable. Like
univariate cases to study bivariate discrete and continuous random variables,
we will need the cumulative distribution function of the random variable (X, Y).
So, let us first define the cumulative distribution function (CDF) of the random
variable (X, Y) in the next section. … (6.1)

6.3 CDF FOR BIVARIATE RANDOM VARIABLE

Recall that to define CDF of a univariate random variable, we used the Borel
measurable sets of the form ( − , x], x  which generated Borel  -field
( ). If you want you may refer to (3.69). In univariate cases, random
variables attain real values so, we needed to define Borel  -field ( ). But in
bivariate cases, random variable (X, Y) attains ordered pair (x, y) of real
values. So, we need to define Borel  -field 2
. Like univariate case the( )
collection 4 = ( − , x]  ( − , y] : x, y   generates Borel  -field ( ). 2

That is  ( 4 ) = ( 2 ). … (6.2)

Like the univariate case, the next concept that is required to define before
defining CDF of a bivariate random variable (X, Y) is the probability measure
X, Y induced by the bivariate random variable (X, Y). Let ( , , ) be a
probability space and X :  → , Y :  → be two random variables then the
probability measure X, Y induced by the bivariate random variable (X, Y) is
defined by (if required you may refer 4.16, 4.27, 4.28, 4.31 and 4.32)

X, Y (B ) = (  : ( X(), Y())  B) , B ( )2

… (6.3)

or X, Y (B ) = ((X, Y)  B ) , B ( )2
… (6.4)

Following the similar steps as we did in Sec. 4.4 of Unit 4 for the univariate
case, we can prove that X, Y is a probability measure on 2
and so ( )
( 2
, ( ),
2
X, Y ) is a probability space and known as induced probability
2
space by the random variable (X, Y) on the xy-plane ( ).

It implies that the probability measure X, Y assigns probability to every

member of ( ) . But we know that (you may refer to 6.2)
2

( 4 )= ( ), where = (− , x]  (− , y] : x, y  
2
4

So, all the rectangles of the form ( − , x]  ( − , y], x, y  are members of

the Borel  -field ( ) . So, by definition of the probability measure
2
X, Y , you
may refer to (6.3), we have

X, Y (( − , x]  ( − , y] ) = (   : ( X(), Y())  (− , x]  (− , y]) ... (6.5)

( − , x]  ( − , y]  ( ) , x, y 
2

137
Probability  X, Y (( − , x]  ( − , y] ) = (  : X()  x, Y()  y) ... (6.6)
Measure and
Random ( − , x]  ( − , y]  ( ),
2
x, y 
Variable
 a  ( − , x]  a  x and b  ( − , y]  b  y 

Like univariate case and due to the frequent use in obtaining probability of the
random variable (X, Y) lying in different intervals, we use the following simple
notation for (6.5) or (6.6).

X, Y ((− , x]  (− , y]) = ( X  x, Y  y) ... (6.7)

( − , x]  ( − , y]  ( ), 2
x, y 

Now, we can define CDF for bivariate random variable (X, Y) as follows.

CDF for Bivariate Random Variable (X, Y): Let ( , , ) be a probability

space and ( 2
, ( ),
2
X, Y ) be the induced probability space induced by the
random variable (X, Y). Then a function F X, Y : 2
→ [0, 1] defined by

F X, Y (x, y) = (( − , x]  ( − , y] ) , (x, y)  2 where

X, Y

 (6.8)
X, Y ( ( − , x]  ( − , y] ) = (  : X()  x, Y()  y)  (x, y )  2 

is called cumulative distribution function (CDF) or simply distribution function

of the bivariate random variable (X, Y) corresponding to the induced
probability measure X, Y .

Using simplified notation used in (6.7), we can write (6.8) as follows

F X, Y (x, y ) = X, Y ((− , x]  (− , y]) = ( X  x, Y  y ) , (x, y)  2
… (6.9)

We will use CDF very frequent so, we will denote it simply by FX, Y instead of
F X, Y . So, using this notation (6.9) can be written as

FX, Y (x, y) = X, Y ((− , x]  (− , y]) = ( X  x, Y  y ) , (x, y)  2

… (6.10)

In view of (6.6), it can also be written as

FX, Y ( x, y) = X, Y ((− , x]  (− , y]) = (  : X()  x, Y()  y) ,(x, y )  2

… (6.11)
Combining (6.8), (6.9), (6.10) and (6.11), we have
F X, Y (x, y) = FX, Y (x, y) = (( − , x]  ( − , y])
X, Y

= (   : X()  x, Y()  y)  … (6.12)

= ( X  x, Y  y ) , (x, y)  2 

Also, we say that FX, Y is the distribution function of the bivariate random
variable (X, Y) instead of saying that corresponding to the induced probability
measure X, Y . … (6.13)
Like univariate case CDF of bivariate random variable (X, Y) also has similar
properties listed as follows.
(a) FX, Y ( − , −  ) = lim FX, Y ( x, y ) = 0. … (6.14)
x →− 
y→ − 
138
 Also, FX, Y ( − , y) = lim FX, Y ( x, y ) = 0 … (6.15) Bivariate Random
x →− 
Variable
and FX, Y (x, −  ) = lim FX, Y ( x, y ) = 0 … (6.16)
y →− 

(b) FX, Y ( ,  ) = lim FX, Y ( x, y ) = 1. … (6.17)

x→ 
y→ 

But FX, Y (x,  ) gives marginal CDF of the random variable X, i.e.,
lim FX, Y (x, y) = FX, Y (x,  ) = FX ( x ) . … (6.18)
y →

Similarly, FX, Y (, y) gives marginal CDF of the random variable Y, i.e.,
lim FX, Y (x, y) = FX, Y (, y) = FY ( y ) . … (6.19)
x →

Marginal CDFs are just CDFs of univariate random variables which you
have studied in Unit 4 and 5. Why the name marginal is added before
CDF to get answer of this question refer to (6.49).

(c) FX, Y is increasing or non-decreasing. That is if x1, x 2 , y1, y 2  be such

that x1  x 2 and y1  y 2 , then FX, Y (x1, y1 )  FX, Y (x 2 , y 2 ) … (6.20)

(d) FX, Y is continuous from above, i.e.,

limFX, Y (x + h, y + k) = FX, Y (x, y)  x, y  … (6.21)

h 0
k 0

(e) If x1, x 2 , y1, y 2  be such that x1  x 2 and y1  y 2 , then (to get more
detail of this property refer 6.63 and Fig. 6.6)

( x1  X  x 2 , y1  Y  y 2 ) = FX, Y (x 2 , y 2 ) − FX, Y (x1, y 2 )

… (6.22)
− FX, Y (x 2 , y1 ) + FX, Y (x1, y1 )

Proofs of these properties are similar to univariate case, so, we are not
discussing them.

6.4 BIVARIATE DISCRETE RANDOM VARIABLE

In Sec. 4.6 of Unit 4, we have defined what is a discrete random variable. Let
( , , ) be a probability space. Here instead of one, we will deal with two
discrete random variables X and Y simultaneously denoted by (X, Y) and
known as discrete bivariate random variable. From the discussion of Sec. 4.6,
you know that a discrete random variable either attains finite number of values
or at the most countably infinite number of values. Here both X and Y are
discrete it means:

• either both X and Y may attain finite number of values or … (6.23)

• both X and Y may attain countably infinite number of values or … (6.24)
• one of them attain finite values and another may attain countably infinite
number of values. … (6.25)

Let us consider one example. Suppose in a bag there are 2 red, 4 blue and 5
black balls. Three balls are drawn from this bag randomly. Let X, Y denote the
number of red and blue balls, respectively out of the three drawn balls. So, X 139
Probability can attain values 0, 1 and 2 while Y can attain values 0, 1, 2 and 3. The
Measure and possible values of the random variable X are shown in the first column of
Random
Variable Table 6.1 and that of Y are shown in the second row of the same table which
is given as follows.

Table 6.1: Values of X and Y together with 12 empty cells where generally
probabilities of corresponding values of the bivariate random
variable (X, Y) are written
Values of Y
Values of X 0 1 2 3
0
1
2
In this table, we have 12 empty cells. We will discuss about the entries of
these 12 empty cells after defining joint probability mass function of bivariate
random variable (X, Y).

In this example, both variables X and Y attain finite number of values. So, the
bivariate random variable (X, Y) is a discrete bivariate random variable.
Possible values of bivariate random variable (X, Y) in this example are (0, 0),
(0, 1), (0, 2), (0, 3), (1, 0), (1, 1), (1, 2), (1, 3), (2, 0), (2, 1), (2, 2), (2, 3). So,
the bivariate random variable (X, Y) attains 12 values. These 12 possible
values are shown in xy-plane in Fig. 6.2 (a) by solid dots. If we write these 12
values in a set S X, Y (say), then S X, Y is known as support of the bivariate
random variable (X, Y), provided if all these 12 points are assigned non zero
probability by probability law. But if one or more than one point among these
12 points have assigned 0 probability by the probability law then that point or
points will go out from the support of (X, Y). So, support of the bivariate
random variable (X, Y) can have at the most 12 points given as follows (to
know general definition of support of (X, Y) you may refer to 6.31) … (6.26)

(0, 0), (0, 1), (0, 2), (0, 3), (1, 0), (1, 1), 
S X, Y =   … (6.27)
(1, 2), (1, 3), (2, 0), (2, 1), (2, 2), (2, 3) 

Fig. 6.2: Visualisation of (a) 12 points which are 12 values of the bivariate random
variable (X, Y) that are possible candidates to form the support of (X, Y) (b) 9

140 points which finally form the support of the bivariate random variable (X, Y)
Why we are calling the bivariate variable (X, Y) is a random variable? We are Bivariate Random
calling it random because it is constituted by random variables X and Y where Variable

both X and Y are defined on a sample space of a random experiment. Then

the immediate next question that will be arising in your mind is if (X, Y) is a
random variable then probabilities should be associated with all the 12
possible values of (X, Y). You are thinking brilliantly and in the right direction.
To answer your brilliant question, first, we have to define what is called joint
probability mass function.

In univariate case in Unit 4, you studied PMF and CDF of a random variable
but in the case of bivariate random variable there are some other concepts
which we need to discuss. So, what we need to discuss in this section are
listed as follows.

• Joint, Marginal and Conditional Probability Mass Functions

• Joint and Marginal Cumulative Distribution Functions

• Independence of Random Variables

Let us discuss these in three subsections of this section as follows.

6.4.1 Joint, Marginal and Conditional Probability Mass

Functions

Joint Probability Mass Function of a Discrete Bivariate Random Variable

Let ( , , ) be a probability space and X, Y be two discrete random
variables on  , then joint probability mass function (JPMF) of the bivariate
random variable (X, Y) is a function which associates unique probability to
each value (x, y) of (X, Y) and 0 probability to the values which are not in the
support of (X, Y). It is denoted by p X, Y and is defined as follows
 ( X = x, Y = y ) , if (x, y) is in support of (X, Y)
p X, Y (x, y) =  … (6.28)
0, otherwise

where ( X = x, Y = y ) = (  : X() = x, Y() = y)  ( x, y)  2

... (6.29)

For a function p X, Y as defined by (6.28) to become a valid joint probability

mass function it has to satisfy the following two conditions.

(i) p X, Y (x, y)  0  (x, y)  support of (X, Y)

(ii) 
( x, y )S X, Y
p X, Y (x, y) = 1 where S X, Y denotes support of (X, Y). … (6.30)

In the definition of joint PMF of the bivariate random variable (X, Y), we have
used the technical word support. We have explained its meaning in (6.26).
Now, as promised there, let us define it in general as follows.
Support of Bivariate Random Variable (X, Y) is the collection of those points
in 2 which have assigned non zero probability by the probability law, i.e.,
Support of (X, Y) is the set S X, Y = (x, y)   2

: p X, Y (x, y)  0 . … (6.31)
141
Probability Let us consider the example of 2 red, 4 blue and 5 black balls considered in
Measure and the beginning of this section. To decide how many of the 12 points will finally
Random
Variable qualify to be members of the support of (X, Y), we have to obtain their
probabilities using (6.28) and (6.29). Let us evaluate the probabilities for each
of the 12 points belong to S X, Y given by (6.27) as follows.

p X, Y (0, 0) = ( X = 0, Y = 0 )
Meaning of ' comma ' in 
= ( getting 0 red ball and 0 blue ball )  
 ( X = 0, Y = 0 ) is ' and' 
= ( All 3 balls are black )
5 54
 
54 54
=   =
3 2 6 1 10
=  =  = ... (6.32)
 11 11  10  9 2 11 10  9 2 165 165
  3
3

Similarly, we can obtain other probabilities as follows.

p X, Y (0, 1) = ( X = 0, Y = 1) = ( getting 0 red ball and 1 blue ball )

 45
  
54
=    = 4
1 2 1 40
 = ... (6.33)
 11 2 165 165
 
3

p X, Y (0, 2) = ( X = 0, Y = 2 ) = ( getting 0 red ball and 2 blue balls )

 45
  
43
=    =
2 1 1 30
5 = ... (6.34)
 11 2 165 165
 
3

p X, Y (0, 3) = ( X = 0, Y = 3 ) = ( getting 0 red ball and 3 blue balls )

 4
 
=   = 4
3 1 4
= ... (6.35)
 11 165 165
 
3

p X, Y (1, 0) = ( X = 1, Y = 0 ) = ( getting 1 red ball and 0 blue ball )

 2 5
  
54
=    = 2
1 2 1 20
 = ... (6.36)
 11 2 165 165
 
3

p X, Y (1, 1) = ( X = 1, Y = 1) = ( getting 1 red ball and 1 blue ball )

 2  4 5
   
=     = 2 4 5
1 1 1 1 40
= ... (6.37)
 11 165 165
 
142 3
p X, Y (1, 2) = ( X = 1, Y = 2 ) = ( getting 1 red ball and 2 blue balls ) Bivariate Random
Variable
 2  4
  
43
=    = 2
1 2 1 12
 = ... (6.38)
 11 2 165 165
 
3

p X, Y (1, 3) = ( X = 1, Y = 3 ) = ( getting 1 red ball and 3 blue balls )

 Only 3 balls are drawn so we  … (6.39)
=0  
cannot get 1 red and 3 blue balls 

p X, Y (2, 0) = ( X = 2, Y = 0 ) = ( getting 2 red balls and 0 blue ball )

 2 5
  
=     = 1 5 
2 1 1 5
= ... (6.40)
 11 165 165
 
3

p X, Y (2, 1) = ( X = 2, Y = 1) = ( getting 2 red balls and 1 blue ball )

 2  4
  
=     = 1 4 
2 1 1 4
= ... (6.41)
 11 165 165
 
3

p X, Y (2, 2) = ( X = 2, Y = 2 ) = ( getting 2 red and 2 blue balls )

 Only 3 balls are drawn so we  … (6.42)
=0  
cannot get 2 red and 2 blue balls 

p X, Y (2, 3) = ( X = 2, Y = 3 ) = ( getting 2 red ball and 3 blue balls )

 Only 3 balls are drawn so we  … (6.43)
=0  
cannot get 2 red and 3 blue balls 

We have obtained probabilities of all 12 points which were lying in S X, Y given

by (6.27). Now, we can complete Table 6.1 by putting these calculated
probabilities in its 12 empty cells. Let us call this new table as Table 6.2.

Table 6.2: Probabilities of all the possible combinations of values of X and Y

Values of Y
Values of X 0 1 2 3
10 40 30 4
0
165 165 165 165
20 40 12 0
1
165 165 165
5 4 0 0
2
165 165

In view of (6.31) three points (1, 3), (2, 2) and (2, 3) will not be part of the
support of the bivariate random variable (X, Y) because they have 0 probability
refer to Table 6.2. So, remaining 9 points form support of the discrete bivariate 143
Probability random variable (X, Y). Joint probability mass function is visualised in Fig. 6.3
Measure and (a) and (b) from two different angles. The height of the line segments at each
Random
Variable of the 9 points of the support represents probabilities of the corresponding
point. … (6.44)

(a)

(b)
Fig. 6.3: Joint probability mass function of bivariate random variable (X, Y) (a) one view
(b) another view

Now, let us define support of individual random variables X and Y in terms of

support of the bivariate random variable (X, Y). If we denote individual
supports of random variables X and Y by S X and S Y respectively then S X
and S Y in terms of support S X, Y of bivariate random variable (X, Y) can be
defined as follows.

S X = x : (x, y)  S X, Y  and SY = y : (x, y)  S X, Y . … (6.45)

Note that Table 6.2 have all information regarding probabilities of all possible
combinations of values of two random variables X and Y. So, this tabular
form is known as joint probability mass of the bivariate random variable
(X, Y). The important point to be noted from here is that if Table 6.2 contains
all information regarding probabilities of all possible combinations of values of
two random variables X and Y then obviously it will contain all information
regarding probabilities of individuals random variables X and Y. This is
discussed next under the heading marginal probability mass functions.

Marginal Probability Mass Functions

The information about probabilities of individual random variables X and Y can

easily be obtained by obtaining rows and columns sums of Table 6.2. After
adding one row indicating columns sums and one column indicating rows
144 sums modified table is shown in Table 6.3 as follows. … (6.46)
Table 6.3: Probabilities of all the possible combinations of values of X and Y Bivariate Random
Variable
with rows and columns sums in the bottom and right margin of the
table

Values of Y
Values of X 0 1 2 3 Rows sums
10 40 30 4 84
0
165 165 165 165 165

20 40 12 0 72
1
165 165 165 165

5 4 0 0 9
2
165 165 165

Columns 35 84 42 4 165
=1
sums 165 165 165 165 165

So, individual probability distribution of random variable X is given by writing

probabilities shown in bold in the last column of the Table 6.3 against different
values of the random variable X which are shown in the first column of the
same Table 6.3. That is, we have to just write the first and the last columns of
Table 6.3 in a separate table say Table 6.4 as follows. … (6.47)

Table 6.4: Marginal probability distribution of the random variable X

Values of X 0 1 2 Total
84 72 9 165
Probabilities ( p X (x) ) =1
165 165 165 165

Similarly, individual probability distribution of random variable Y is given by

writing probabilities shown in bold in the last row of the Table 6.3 against
different values of the random variable Y which are shown in the second row
of the same Table 6.3. That is, we have to just write the second and the last
rows of Table 6.3 in a separate table say Table 6.5 as follows. … (6.48)

Table 6.5: Marginal probability distribution of the random variable Y

Values of Y 0 1 2 3 Total
35 84 42 4 165
Probabilities ( p Y (y) ) =1
165 165 165 165 165

Individual probability distributions of random variables X and Y given by Table

6.4 and Table 6.5, respectively are obtained from joint probability mass
function of the bivariate random variable (X, Y) so they have the special name
called marginal probability mass functions of X and Y, respectively. The
word marginal is used because probabilities of individual random variables X
and Y were lying in the margins (last column and last row) of the Table 6.3.
Marginal probability mass functions of random variables X and Y are
visualised in Fig. 6.4 (a) and (b) respectively. … (6.49)

With the help of this example, you have understood the idea of marginal
probability mass function of individual random variables X and Y. Now, let us
define them in general as follows. 145
Probability
Measure and
Random
Variable

Fig. 6.4: Marginal probability mass functions of random variable (a) X (b) Y

Marginal Probability Mass Functions of X and Y: Let ( , , ) be the

probability space and X, Y be two discrete random variables having values
x1, x 2 , x 3 , , xm and y1, y 2 , y 3 , , yn respectively. If joint probability mass
function of the bivariate random variable (X, Y) is given by

 ( X = x, Y = y ) , if (x, y) is in support of (X, Y)

p X, Y (x, y) = 
0, otherwise

then marginal probability mass function of the random variable X is given by

p X (x) = ( Y = y1 ) ( X = x | Y = y1 ) + ( Y = y 2 ) ( X = x | Y = y 2 ) +
+ ( Y = yn ) ( X = x | Y = y n ) Using total law of probability 
= ( X = x, Y = y1 ) + ( X = x, Y = y 2 ) + + ( X = x, Y = yn )

We know that (E | F ) =
(E  F )  E  F = F E | F 
 ( ) ( ) ( )
 (F ) 

( X = x, Y = y )
n
 p X (x) =  j … (6.50)
j =1

Similarly, marginal probability mass function of the random variable Y is given

by
p Y (y) = ( X = x1 ) ( Y = y | X = x1 ) + ( X = x 2 ) ( Y = y | X = x 2 ) +
+ ( X = xm ) ( Y = y | X = xm ) Using total law of probability 
= ( Y = y, X = x1 ) + ( Y = y, X = x 2 ) + + ( Y = y, X = xm )

(X = x , Y = y)
m
 p Y (y) =  i … (6.51)
i =1

So, (6.50) and (6.51) define marginal probability mass functions of X and Y
respectively in general.

Now, we define conditional probability mass functions as follows.

Conditional Probability Mass Functions

In Sec. 1.6 of Unit 1, we have learnt to calculate conditional probability of an

event E (say) for some given event F (say). Recall that conditional probability
of event E given event F is denoted by (E | F ) . Also, recall that if events E
146 and F are related to the sample space  then sample space for the event
E | F reduces from  to F. That is in condition probability conditioning event is Bivariate Random
Variable
the reduced sample space. … (6.52)

Recall one more thing from Unit 4, if X is a random variable defined on the
sample space  then X = x is an event G (say) where

G =    : X() = x … (6.53)

Now, you can understand conditional probability mass function easily. Let (X,
Y) be a discrete bivariate random variable on the sample space  where X
and Y can attain values x1, x 2 , x 3 , , xm and y1, y 2 , y 3 , , yn , respectively. If
p X, Y (x, y) denotes joint PMF of (X, Y) and p X (x), pY (y) be marginal
probability mass functions of random variables X and Y respectively, then
conditional probability mass function of the random variable X for given Y = y j
is denoted by p X|Y (x | y j ) and is defined by

p(X = x, Y = y j )
p X|Y (x | y j ) = , p Y (y j )  0, x = x1, x 2 , x 3 , , xm ... (6.54)
p Y (Y = y j )
 (E  F ) . You may refer to (1.28). Here 
 (E | F ) = 
 (F ) 
 
E =   : X() = x and F =   : Y() = y j  

For a valid conditional probability mass function following two conditions

should be satisfied by p X|Y (x | y j ) :

(i) 0  p X|Y (x | y j )  1  x = x1, x 2 , x 3 , , xm … (6.55)

m
(ii) p
i =1
X|Y (x i | y j ) = 1 … (6.56)

Similarly, conditional probability mass function of the random variable Y for

given X = x i is denoted by p Y| X (y | x i ) and is defined by

p(X = x i , Y = y)
p Y| X (y | xi ) = , p X (x i )  0, y = y1, y 2 , y 3 , , yn ... (6.57)
p X (X = x i )

For a valid conditional probability mass function following two conditions

should be satisfied by p Y| X (y | x i ) :

(i) 0  p Y| X (y | xi )  1  y = y1, y 2 , y 3 , , ym … (6.58)

n
(ii) p
j =1
Y| X (y j | x i ) = 1 … (6.59)

For example, in the example of 2 red, 4 blue and 5 black balls if Y = y j = 1,

then conditional probability mass function of the random variable X for given
Y = y j = 1 is given as follows and shown in Fig. 6.5 (a).

p(X = x, Y = 1)
p X|Y (x | y = 1) = , x = 0, 1, 2
p Y ( Y = 1)
147
Probability   p(X = 0, Y = 1) 40 / 165 40 
Measure and 40 / 84, x = 0  p X|Y (X = 0 | Y = 1) = = = 
Random   p Y (Y = 1) 84 / 165 84 
 
Variable  p(X = 1, Y = 1) 40 / 165 40 
= 40 / 84, x = 1 p X|Y (X = 1| Y = 1) = = =  (6.60)
  p Y (Y = 1) 84 / 165 84 

4 / 84, x = 2  p ( X = 2 | Y = 1) = p(X = 2, Y = 1) = 4 / 165 = 4 
 
X|Y
p Y ( Y = 1) 84 / 165 84 

Similarly, conditional probability mass function of the random variable Y for

given X = xi = 1 is given as follows and shown in Fig. 6.5 (b).
p(Y = y, X = 1)
p Y| X (y | x = 1) = , y = 0, 1, 2, 3
p X (X = 1)
  p(Y = 0, X = 1) 20 / 165 20 
20 / 72, y = 0  p Y| X (Y = 0 | X = 1) = = = 
  p X (X = 1) 72 / 165 72 

40 / 72, y = 1  p Y| X (Y = 1| X = 1) =
p(Y = 1, X = 1) 40 / 165 40 
= = 
  p X (X = 1) 72 / 165 72 
= (6.61)
  p(Y = 2, X = 1) 12 / 165 12 
12 / 72, y = 2  p Y| X (Y = 2 | X = 1) =
p X (X = 1)
= = 
  72 / 165 72 
  p(Y = 3, X = 1) 0 
0, y=3  p Y| X (Y = 3 | X = 1) = = = 0
  p X (X = 1) 72 / 165 

Fig. 6.5: Conditional probability mass functions of random variable (a) X given Y = 1 (b) Y
given X = 1

6.4.2 Joint and Marginal Cumulative Distribution

Functions
Joint Cumulative Distribution Function
Joint CDF has already been defined (you may refer to 6.12).
Keep following important point in mind. Recall that in the case of univariate
random variable refer to (4.36d), we have
( a  X  b ) = FX (b) − FX (a) … (6.62)

But in bivariate case if a, b, c, d  be such that a  b and c  d, then

probability of the type (6.62) of the joint random variable (X, Y), i.e.,
( a  X  b,c  Y  d) using joint CDF of random variables X and Y is given by
148 (a  X  b, c  Y  d) = FX, Y (b, d) − FX, Y (a, d) − FX, Y (b, c) + FX, Y (a, c) … (6.63)
Let us visualise rectangular regions corresponding to each term of (6.63) in Bivariate Random
Fig. 6.6 (a1) to (a5) and specify them as follows. Variable

• LHS of (6.63): The rectangular region corresponding to LHS of (6.63) is

shown by rectangle ABCD in Fig. 6.6 (a1).

• First Term of RHS of (6.63): The rectangular region corresponding to the

first term of RHS of (6.63) is shown by the rectangle APRT in Fig. 6.6 (a2).

• Second Term of RHS of (6.63): The rectangular region corresponding to

the second term of RHS of (6.63) is shown by the rectangle BPRS in Fig.
6.6 (a3).

• Third Term of RHS of (6.63): The rectangular region corresponding to the

third term of RHS of (6.63) is shown by the rectangle DQRT in Fig. 6.6 (a4).

• Fourth Term of RHS of (6.63): The rectangular region corresponding to

the fourth term of RHS of (6.63) is shown by the rectangle CQRS in Fig. 6.6
(a5).
Due to the importance of the result given by (6.63), we have visualised the
idea behind each of its term. We will use this result in solving problems. So,
make good understanding of this result and keep it in your mind.
Marginal Cumulative Distribution Functions
Let ( , , ) be the probability space. Let x1, x 2 , x 3 , , xm and
y1, y 2 , y 3 , , yn be the supports of random variables X and Y respectively. Let
us define events A and Ei , 1, 2, 3, , n as follows.

A =   : X()  x, x = x1, x 2 , x3 , , xm and

E j =   : Y() = y j , j = 1, 2, 3, ,n

Since y1, y 2 , y 3 , , yn form support of the random variable Y, so

(E )  0, 
j j = 1, 2, 3, , n.

So, using total law of probability, marginal CDF of the random variable X is
given by
( X  x ) = ( A ) =  (E j ) ( A | E j )
n
FX (x ) =
j =1

(Y = y ) (X  x | Y = y )
n
= j j
j =1

 (E  F )  F E | F = E  F 
( X  x, Y = y ) 
n
= (E | F ) = ( ) ( ) ( )
j =1
j
 (F ) 

 FX (x) = FX (x, ) By definition of CDF ... (6.64)

or FX (x) = lim FX (x, y)

y →
By definition of CDF ... (6.65)

Similarly, marginal CDF of the random variable Y is given by

FY (y) = FX, Y ( , y) ... (6.66)
or FY (y) = lim FX, Y (x, y) ... (6.67)
x →
149
Probability
Measure and
Random
Variable

Fig. 6.6: Visualisation of different terms of (6.63) (a) LHS (b) first term of RHS (c) second
term of RHS (d) third term of RHS (e) fourth term of RHS

6.4.3 Independence of Random Variables

Let ( , , ) be the probability space. Let x1, x 2 , x 3 , , xm and

y1, y 2 , y 3 , , yn be the supports of random variables X and Y respectively.

Here, we want to discuss independence of two random variables. Recall that

in Sec. 1.8 of Unit 1, we have learnt when two or more events are said to
independent and what is the condition for independence, you may refer to
150 (1.45) and (1.47) to (1.49). So, it will help you in understanding the idea of
independence of two random variables if we connect independence of random Bivariate Random
variables X and Y with the independence of events discussed in Unit 1. Variable

Independence of two random variables can be expressed in three equivalent

ways mentioned as follows.
• Independence in terms of CDF
• Independence in terms of PMF
• Independence in terms of conditional PMF
Let us discuss these taken one at a time.

• Independence in Terms of CDF

As mentioned earlier, we will connect independence of random variables with
independence of events. Keeping this in view, let us define events E and F as
follows.

E =   : X()  x , x = x1, x 2 , x 3 , , xm and

… (6.68)
F =   : Y()  y , y = y1, y 2 , y 3 , , yn

We know that (you may refer to 1.47) if events E and F are independent under
the probability measure then

(E  F ) = (E ) (F ) … (6.69)

Now, if events E and F are independent for all possible values of x and y then
FX, Y (x, y) = ( X  x, Y  y) Using definition of CDF refer to (6.10)

= (E  F ) Using (6.68)

= (E ) (F ) Using (6.69)

= (X  x) (Y  y) Using (6.68)

= FX (x)FX (x) Using (4.36a)

So, random variables X and Y are independent if following holds.

FX, Y (x, y) = FX (x)FY (y)  x  S X and y  S Y ... (6.70)

• Independence in Terms of PMF

Equation (6.70) expresses condition of independence in terms of CDF.

Similarly, we can express condition of independence in terms of PMF instead
of CDF if we define events E and F as follows.

E =   : X() = x , x = x1, x 2 , x 3 , , xm and

F =   : Y() = y , y = y1, y 2 , y 3 , , yn

Following similar steps, we can prove that random variables X and Y will be
independent if

p X, Y (x, y) = p X (x)p X (x)  x  S X and y  S Y ... (6.71)

or ( X = x, Y = y) = (X = x) (Y = y)  x  S X and y  S Y ... (6.72)

151
Probability • Independence in Terms of Conditional PMF
Measure and
Random You have studied conditional PMF in this section refer (6.52) to (6.61). Two
Variable
random variables X and Y are said to be independent if conditional PMF’s
satisfy the following conditions
p X|Y (x | y) = p X (x)  x  S X , y  S Y and p Y (y)  0 ... (6.73)

or p Y| X (y | x) = p Y (y)  x  S X , y  S Y and p X (x)  0 ... (6.74)

Now, we discuss one example to apply different ideas discussed in this

section.

Example 1: Suppose we have a binary communication channel where support

of both input and output random variables X and Y is {0, 1}. Input random
variable X attains values 0 and 1 in equally likely fashion. But channel noise is
a common problem with a communication channel due to which input 0 may
be transmitted to 1 and similarly input 1 may be transmitted to 0. We are given
the following channel transition probabilities:

( Y = 0 | X = 0 ) = 0.99 and ( Y = 1| X = 1) = 0.92. … (6.75)

On the basis of this information obtain:

(a) joint probability mass function of (X, Y).
(b) marginal PMF’s of X and Y.
(c) joint CDF of (X, Y).
(d) marginal CDF’s of X and Y.
(e) conditional PMF of Y given X = 1.
(f) are random variables X and Y independent?

Solution: Since input values 0 and 1 are equally likely and X represents input
random variable, so, we have
( X = 0 ) = 0.5 and ( X = 1) = 0.5. … (6.76)

Also, if input is 0 then either output will be 0 or 1, so, in conditional space, we

should have

( Y = 0 | X = 0 ) + ( Y = 1| X = 0 ) = 1
 ( Y = 1| X = 0 ) = 1 − ( Y = 0 | X = 0 )
= 1 − 0.99 Using (6.75)

= 0.01 ... (6.77)

Similarly, ( Y = 0 | X = 1) = 1 − ( Y = 1| X = 1) = 1 − 0.92 = 0.08 … (6.78)

Model input values, transition probabilities, output values and probabilities

obtained in (6.77) and (6.78) all are shown in Fig. 6.7.
(a) Here random variables X and Y both attain values 0 and 1. Since
random variables X and Y both assume finite number of values (2 in
this case), so (X, Y) is a discrete bivariate random variable. To obtain
152 joint PMF of the
 Bivariate Random
Variable

Fig. 6.7: Visualisation of the communication channel with transition probabilities, input
values and output values

discrete bivariate random variable (X, Y), we have to fill up four empty
cells of the Table 6.6 with corresponding probabilities.
Table 6.6: Number of cells formed by possible values of X and Y

Values of Y
Values of X 0 1
0
1
Let us obtain probabilities of these four cells as follows.
( X = 0, Y = 0 ) = ( X = 0 ) ( Y = 0 | X = 0 )
= 0.5  0.99 Using (6.76) and (6.75)
= 0.495
Similarly, ( X = 0, Y = 1) = ( X = 0 ) ( Y = 1| X = 0 )
= 0.5  0.01 = 0.005 Using (6.76) and (6.77)
( X = 1, Y = 0 ) = ( X = 1) ( Y = 0 | X = 1)
= 0.5  0.08 = 0.04 Using (6.76) and (6.78)
( X = 1, Y = 1) = ( X = 1) ( Y = 1| X = 1)
= 0.5  0.92 = 0.46 Using (6.76) and (6.75)

We have obtained probabilities of the four cells of the Table 6.6 and are
shown in the new Table 6.7. The tabular form given by Table 6.7 is known
as joint PMF of the random variable (X, Y).
Table 6.7: Joint probability mass function of discrete bivariate random
variable (X, Y)

Values of Y
Values of X 0 1
0 0.495 0.005
1 0.04 0.46
153
Probability (b) To obtain marginal PMF’s of the random variables X and Y, first, we have
Measure and to find out sum of each row and column of Table 6.7. Rows and columns
Random
Variable sums are shown in Table 6.8.
Table 6.8: Marginal probability mass functions of discrete random variables
X and Y as sums of each row and column

Values of Y
Values of X 0 1 Marginal probabilities
0 0.495 0.005 0.5
1 0.04 0.46 0.5
Marginal probabilities 0.535 0.465 1

So, marginal probability mass function of the random variable X is given

by Table 6.9 as follows.
Table 6.9: Marginal probability mass function of the random variable X

Values of X 0 1 Total
Probabilities ( p X (x) ) 0.5 0.5 1

Similarly, marginal probability mass function of the random variable Y is

given by in Table 6.10 as follows
Table 6.10: Marginal probability mass function of the random variable Y

Values of Y 0 1 Total
Probabilities ( p Y (y) ) 0.535 0.465 1
(c) For finding joint CDF of (X, Y), we have to compute FX, Y (x, y) for different
values of x and y using (6.10). So, joint CDF of discrete bivariate random
variable (X, Y) is given by Table 6.11.
Table 6.11: Joint CDF of discrete bivariate random variable (X, Y)

Values of Y
Values of X 0 1
0 0.495 0.5
1 0.535 1

 FX, Y (0, 0) = ( X  0, Y  0 ) = ( X = 0, Y = 0 ) = 0.495 

 
FX, Y (0, 1) = ( X  0, Y  1) = ( X = 0, Y = 0 ) + ( X = 0, Y = 1) 
 
 = 0.495 + 0.005 = 0.5, etc. 

(d) Marginal CDF of random variable X is given by combining values of X

from the first column and corresponding probabilities from the third
column of Table 6.11 and is given by Table 6.12.
Table 6.12: Marginal CDF of the random variable X

Values of X 0 1
Probabilities ( p X (x) ) 0.5 1

Marginal CDF of random variable Y is given by combining values of Y

from the first row and corresponding probabilities from the third row of
154 Table 6.11 and is given by Table 6.13.
 Table 6.13: Marginal CDF of the random variable Y Bivariate Random
Variable
Values of Y 0 1
Probabilities ( p Y (y) ) 0.535 1

(e) We can obtain conditional PMF of random variable Y for given X = 1 as

follows.
p X, Y (X = 1, Y = y)
p Y| X (y | X = 1) = , y = 0, 1
p X (X = 1)
  p X, Y (X = 1, Y = 0) 0.04 
0.08, y = 0  p Y| X (Y = 0 | X = 1) = = = 0.08 
  p X (X = 1) 0.5 
=
  p X, Y (X = 1, Y = 1) 0.46 
0.92, y = 1  p Y| X (Y = 1| X = 1) = p X (X = 1)
= = 0.92 
  0.5 

(f) ( X = 0, Y = 0 ) = 0.495 ... (6.79)

( X = 0 ) ( Y = 0 ) = 0.5  0.535 = 0.2675 ... (6.80)

From (6.79) and (6.80) ( X = 0, Y = 0 )  ( X = 0 ) ( Y = 0 )

Since (6.72) does not hold so random variables X and Y are not
independent.

Now, you can try the following Self-Assessment Question.

SAQ 1

Joint PMF is given and find out other things: Joint PMF of a discrete
bivariate random variable (X, Y) is given by
k(2x + 3y), x = 0, 1, 2; y = 1, 2, 3, 4
p X, Y (x, y) = 
0, otherwise

(a) Find value of k.

(b) Find the joint PMF of (X, Y) in formula form and tabular form. Also find
marginal PMF’s of X and Y.
(c) Find joint CDF of (X, Y).
(d) Find marginal CDF’s of X and Y.
(e) Find conditional PMF of X given Y = 3.
(f) Are X and Y independent?
(g) Find ( 0  X  2, 1  Y  3 ) .

6.5 BIVARIATE CONTINUOUS RANDOM

VARIABLE
In the previous section, you have studied discrete bivariate random variable
and many concepts related to that like joint, marginal and conditional PMF’s;
joint and marginal CDF’s; and finally, independence. In this section, we will
study such concepts for continuous bivariate random variable. But in Unit 5, 155
Probability you have already studied univariate continuous random variable. So, from Unit
Measure and 5, you know how to handle things related to a continuous random variable in
Random
Variable continuous world. Combining the knowledge of Unit 5 and previous section of
this unit, you can easily understand the corresponding concepts for continuous
bivariate random variable. So, we are not going to explain them in detail like
discrete bivariate random variable in the previous section. Here, we will define
the terms related to continuous bivariate random variable in brief as follows
and after that we do one example to explain them.
Joint Cumulative Distribution Function
Joint CDF has been already defined refer to (6.12).
Recall that in the case of discrete bivariate random variable refer to (6.63), we
have
(a  X  b, c  Y  d) = FX, Y (b, d) − FX, Y (a, d) − FX, Y (b, c) + FX, Y (a, c) … (6.81)

But like univariate case in continuous world a single point has zero probability,
i.e.,
( X = a, Y = b) = 0 … (6.82)

In view of (6.82), we can say that all expressions like

( a  X  b, c  Y  d) , (a  X  b, c  Y  d) , (a  X  b, c  Y  d ),
( a  X  b, c  Y  d) , (a  X  b, c  Y  d), (a  X  b, c  Y  d ) , etc.
all are equal to FX, Y (b, d) − FX, Y (a, d) − FX, Y (b, c) + FX, Y (a, c). … (6.83)

Marginal Cumulative Distribution Functions

Marginal CDF of the random variable X is given by
FX (x) = ( X  x ) = FX (x, ) = yli→
m FX (x, y) ... (6.84)

Similarly, marginal CDF of the random variable Y is given by

FY (y) = FX, Y (, y) = lim FX, Y (x, y) ... (6.85)
x →

Joint Probability Density Function of a Jointly Continuous Bivariate

Random Variable

Let (X, Y) be a jointly continuous random variable. The jointly continuous

random variable (X, Y) will have a joint probability density function (JPDF) fX, Y
if there exists a Borel measurable function fX, Y :  → [0, ] such that
 

X, Y ( B ) = ((X, Y)  B ) =   IB fX, Y (x, y) dx dy, B ( ) ( )

− − … (6.86)
where IB is indicator function of B

In particular, if B = (a, b]  (c, d] , then

d b

X, Y ( (a, b]  (c, d] ) = ((X, Y)  (a, b]  (c, d] ) =   fX, Y (x, y) dx dy … (6.87)

c a

In Secs. 9.4 and 9.5 of Unit 9 of the course MST-011, you have studied to
156 evaluate double integral. You also know that double integral shown in (6.87)
gives volume of the region bounded by five planes x = a, x = b, y = c, y = d, Bivariate Random
z = 0 and one surface given by z = fX, Y (x, y). But in the world of probability Variable

theory double integral shown in (6.87) gives the probability

((X, Y)  (a, b]  (c, d] ) . Therefore, to be a valid joint PDF it must satisfy the
following two conditions.

 Probabilities are always non-negative 

(a) fX, Y (x, y)  0  (x, y)    density shoud be  0 
 
… (6.88)
 
 Sum of all probabilities should be 1 and in 
(b)  I f
− −
B X, Y ( x, y )dx dy = 1 
continuous world sum is given by integration


… (6.89)
Keeping (6.87) in view, (6.10) can be written as follows.
y x
FX, Y ( x, y ) = ((X, Y)  ( − , x]  ( − , y] ) =   fX, Y (x, y) dx dy … (6.90)
− −

Marginal Probability Density Functions

If fX, Y (x, y) be the joint PDF of jointly continuous random variable (X, Y) then
marginal probability density function of X and Y, respectively, are given by
 
fX ( x ) =  fX, Y (x, y) dy and fY ( y ) = f X, Y (x, y ) dx … (6.91)
− −

Conditional Probability Density Functions

If fX, Y (x, y) be the joint PDF of jointly continuous random variable (X, Y) and
fX ( x ) , fY ( y ) be marginal PDF’s of X and Y respectively, then

• Conditional PDF of X given Y is given by

fX, Y (x, y)
fX|Y ( x | y ) = , fY ( y )  0 and … (6.92)
fY ( y )

• Conditional PDF of Y given X is given by

fX, Y (x, y)
fY|X ( y | x ) = , fX ( x )  0 … (6.93)
fX ( x )

Independence of Random Variables

Like discrete bivariate case, independence of two jointly continuous random

variables X and Y can be expressed in three ways.

• Independence in terms of CDF

• Independence in terms of PDF

• Independence in terms of conditional PDF

Let us discuss these taken one at a time.

157
Probability • Independence in Terms of CDF
Measure and
Random Jointly continuous random variables X and Y are independent if the following
Variable holds.
FX, Y (x, y) = FX (x)FY (y)  x, y  … (6.94)

• Independence in Terms of PDF

Jointly continuous random variables X and Y are independent if the following
holds.
fX, Y (x, y) = fX (x) fY (y)  x, y  … (6.95)

• Independence in Terms of Conditional PDF

Jointly continuous random variables X and Y are independent if the following
holds.
fX|Y ( x | y ) = fX ( x )  x, y  and fY ( y )  0 … (6.96)

or fY|X ( y | x ) = fY ( y )  x, y  and fX (x)  0 … (6.97)

Now, we discuss one example to apply different ideas discussed in this

section.
Example 2: Joint PDF of jointly continuous random variable (X, Y) is given by

k(4 − x)(5 − y), 0  x  4, 0  y  5

fX, Y (x, y) =  … (6.98)
0, otherwise

Obtain:
(a) value of k.
(b) marginal PDF’s of X and Y.
(c) joint CDF of (X, Y).
(d) marginal CDF’s of X and Y.
(e) conditional PDF of X given Y.
(f) are random variables X and Y independent?

Solution: (a) We can find the value of k using (6.89) k as follows.

  5 4

  IB fX, Y (x, y) dx dy = 1   k(4 − x)(5 − y) dx dy = 1

− − 0 0
x=4
5
 x2 
5
  k(5 − y)  4x −  dy = 1   k(5 − y) 16 − 8  dy = 1
0  2  x =0 0
y =5
 y2   25  1
 8k 5y −  = 1  8k 25 −  = 1  100k = 1  k = 100 ... (6.99)
 2  y =0  2 

(b) We can find marginal PDF of X using (6.91) as follows.

 y =5
4−x 4−x y2 
5
fX ( x ) =  fX, Y (x, y) dy =
100 0
(5 − y) dy =  5y − 
−
100  2  y =0
4−x 25  4 − x
=  25 − = , 0x4 ... (6.100)
158 100  2  8
 Similarly, marginal PDF of Y is given by Bivariate Random
Variable
 x=4
5−y 5−y x2 
4
f Y ( y ) =  fX, Y (x, y) dx =
100 0
(4 − x) dx =  4x − 
−
100  2  x =0
5−y 10 − 2y
=
100
 16 − 8 =
25
, 0y5 ... (6.101)

(c) Using (6.90), joint CDF is given by

y x
FX, Y ( x, y ) =  f X, Y (x, y) dx dy … (6.102)
− −

To evaluate (6.102), we generally consider five cases I, II, III, IV and V as

shown in Fig. 6.8 (a) for the general case where a  x  b and c  y  d.
In the present case, we have a = 0, b = 4, c = 0, d = 5. So, we have to
consider the following five cases.

Case I: x < 0 or y < 0

In this case in view of (6.98) and (6.102), we have FX, Y ( x, y ) = 0.

Case II: x  4 and y  5

In this case, we have FX, Y ( x, y ) = 1  Sum of all probability is 1

Case III: 0  x  4 and 0  y  5

(5 − y)(4 − x)
y x y x

FX, Y ( x, y ) =  fX, Y (x, y) dx dy =   dx dy

− − 0 0
100
x
(5 − y)  x2  (5 − y)  x2 
y y

=  4x −  dy = 0 100  4x −  dy
0
100  2 0 2

=
1  x2   y2  8x − x 2 10y − y 2
y
( )( )
 4x −  5y −  = ... (6.103 )
100  2  2 0 400

Case IV: 0  x  4 and y  5

FX, Y ( x, y ) = FX, Y ( x, 5 )

( 8x − x )(10  5 − 5 )
2 2

=
400
Putting y = 5 in (6.103)

8x − x 2
= ... (6.104)
16
Case V: x  4 and 0  y  5

FX, Y ( x, y ) = FX, Y ( 4, y )

( 8  4 − 4 )(10y − y )
2 2

=
400
Putting x = 4 in (6.103)

10y − y 2
= ... (6.105)
25

Hence, on combining all cases CDF of (X, Y) is given by

159
Probability 0, if x  0 or y  0

Measure and
Random (
 8x − x 10y − y
2 2
)( )
, 0  x  4, 0  y  5
Variable  400

 8x − x 2
FX, Y (x, y) =  , 0  x  4, y  5 … (6.106)
 16
10y − y 2
 , x  4, 0  y  5
 25
1, if x  4 and y  5

(d) Combining (6.84) and (6.104), marginal CDF of X is given by
0, if x  0

 8x − x
2
FX (x) =  , 0x4
 16
1, if x  4

Similarly, combining (6.85) and (6.105) marginal CDF of Y is given by

0, if y  0

10y − y
2
FY (y) =  , 0y5
 25
1, if y  5

(e) Using (6.92), conditional density of X given Y is given by

fX, Y (x, y)
fX|Y ( x | y ) = , fY ( y )  0
fY ( y )
( 8x − x )(10y − y )
2 2

= 400
10y − y 2
Using (6.98), (6.99) and (6.101)
25
8x − x 2
= , 0x4
16

(f) Since FX, Y (x, y) = FX (x)FY (y)  x, y 

Hence, random variables X and Y are independent.

Fig. 6.8: Visualisation of the (a) five regions corresponding to five cases in part (c) of
Example 2 (b) vertical strip PQ to obtain limits of y in terms of x refer solution
160 of SAQ 3
In this example, joint PDF was given and we obtained joint CDF using the Bivariate Random
integral given by (6.90). But you know that differentiation and integration are Variable

reverse operators so if joint CDF is given then joint PDF can be obtained by
partially differentiating CDF with respect to x and y, i.e.,

2 2
fX, Y (x, y) =
xy
( FX, Y (x, y) ) or fX, Y (x, y) =
yx
(FX, Y (x, y) ) ... (6.107)

 2 2 
Here we are assuming that
xy
( FX, Y (x, y) ) =
yx
( FX, Y (x, y) ) . But in 
 
general it does not hold. In fact, it holds only when first order partial 
 
derivatives  (F (x, y) ) and  (F (x, y) ) are continuous functions 
 x
X, Y
x
X, Y

 

You will study partial derivative in the course MST-022 in detail. Let us briefly
explain it through an example as follows. … (6.107a)
2
xy
(F ( x, y ) ) means first differentiate F ( x, y ) partially w.r.t. y and then the
resulting function w.r.t. x. When we differentiate a function partially w.r.t. one
variable, then the other variable is treated as constant … (6.107b)
For example, Let F ( x, y ) = x 5 y 3 + x13 y 7

If we differentiate it partially w.r.t. y, we have

 Here, x is treated
y
( ( ) ( )
F ( x, y ) ) = x 5 3y 2 + x13 7y 6 = 3 x 5 y 2 + 7x13 y 6  
as constant 

If we now partially differentiate this resulting expression w.r.t. x, we have

2 Here, y is treated
x y
( ( ) ( )
F ( x, y ) ) = 3 5x 4 y 2 + 7 13x12 y 6 = 15 x 4 y 2 + 91x12 y 6  
as constant 
If we first partially differentiate w.r.t. x and then w.r.t. y, we get
 Here, y is treated
x
( ( )
F ( x, y ) ) = 5x 4 y 3 + 13x12 y 7 = 5 x 4 y 3 + 13x12 y 7  
as constant 

2 Here, x is treated
y x
( ( ) ( )
F ( x, y ) ) = 5x 4 3 y 2 + 13x12 7y 6 = 15 x 4 y 2 + 91x12 y 6  
as constant 

Note that here both way we get the same result.

Now, you can try the following two Self-Assessment Questions.

SAQ 2

For given Joint CDF Obtain Joint PDF: In Example 2 joint PDF was given
and you obtained joint CDF given by (6.106). Here assume that joint CDF is
given to you. Find joint PDF of the jointly continuous random variable (X, Y).
SAQ 3

For given CDF obtain PDF and probabilities: Consider the same jointly
continuous random variable (X, Y) discussed in Example 2. Find
( X + Y  2) .
161
Probability 6.6 SUMMARY
Measure and
Random
Variable A brief summary of what we have covered in this unit is given as follows:
• Bivariate Random Variable: Let X :  → and Y :  → be two random
variables. The ordered pair random variable (X, Y) which associates a
unique ordered pair of real numbers to each member of the sample space
( ) is called a bivariate random variable on  .

• If both X and Y are discrete random variables then the ordered pair random
variable (X, Y) is called a discrete bivariate random variable. If both X
and Y are jointly continuous random variables then the ordered pair
random variable (X, Y) is called a continuous bivariate random variable.

• CDF of (X, Y) is defined by

F X, Y (x, y ) = X, Y ((− , x]  (− , y]) = ( X  x, Y  y ) , (x, y)  2

• Properties of CDF are listed as follows:

• FX, Y ( − , −  ) = lim FX, Y ( x, y ) = 0

x →− 
y→ − 

• FX, Y ( − , y) = lim FX, Y ( x, y ) = 0

x →− 

• FX, Y (x, −  ) = lim FX, Y ( x, y ) = 0

y →− 

• FX, Y ( ,  ) = lim FX, Y ( x, y ) = 1

x→ 
y→ 

• Marginal CDF of X is lim FX, Y (x, y) = FX, Y (x,  ) = FX ( x ) .

y →

• Marginal CDF of Y is lim FX, Y (x, y) = FX, Y (, y) = FY ( y ) .

x →

• FX, Y is increasing or non-decreasing. That is if x1, x 2 , y1, y 2  be such

that x1  x 2 and y1  y 2 , then FX, Y (x1, y1 )  FX, Y (x 2 , y 2 ) .

• FX, Y is continuous from above, i.e.,

limFX, Y (x + h, y + k) = FX, Y (x, y)  x, y  .

h 0
k 0

• If x1, x 2 , y1, y 2  be such that x1  x 2 and y1  y 2 , then (to get more

detail of this property refer to 6.63 and Fig. 6.6)

( x1  X  x 2 , y1  Y  y 2 ) = FX, Y (x 2 , y 2 ) − FX, Y (x1, y 2 )

− FX, Y (x 2 , y1 ) + FX, Y (x1, y1 )

 ( X = x, Y = y ) , if (x, y) is in support of (X, Y)

• Joint PMF: p X, Y (x, y) = 
0, otherwise

• Marginal probability mass function of the random variable X is given by

( X = x, Y = y )
n
p X (x) =  j
j =1
162
 ( X = x , Y = y)
m
Bivariate Random
Similarly, marginal PMF of Y is given by p Y (y) =  i Variable
i =1

• Conditional PMF of X given Y is

p(X = x, Y = y j )
p X|Y (x | y j ) = , p Y (y j )  0, x = x1, x 2 , x3 , , xm
pY ( Y = y j )

Similarly, conditional PMF of Y given X is

p(X = x i , Y = y)
p Y| X (y | xi ) = , p X (x i )  0, y = y1, y 2 , y 3 , , yn
p X (X = x i )

• Two random variables X and Y are said to be independent if

• For both discrete and continuous case
FX, Y (x, y) = FX (x)FY (y)  x  S X and y  S Y

• For discrete case

p X, Y (x, y) = p X (x)p X (x)  x  S X and y  S Y

( X = x, Y = y) = ( X = x) (Y = y)  x  S X and y  S Y
p X|Y (x | y) = p X (x)  x  S X , y  S Y and p Y (y)  0
pY|X (y | x) = pY (y)  x  S X , y  S Y and p X (x)  0

• For continuous case

fX, Y (x, y) = fX (x) fY (y)  x, y 

fX|Y ( x | y ) = fX ( x )  x, y  and fY ( y )  0
fY|X ( y | x ) = fY ( y )  x, y  and fX (x)  0

• Probability in terms of joint density

 

X, Y ( B ) = ((X, Y)  B ) =   IB fX, Y (x, y) dx dy, B ( ) ( )

− −

where IB is indicator function of B

• CDF in continuous case

y x
FX, Y ( x, y ) = ((X, Y)  ( − , x]  ( − , y] ) =   fX, Y (x, y) dx dy
− −

• Marginal PDF in continuous case

 
fX ( x ) =  fX, Y (x, y) dy and fY ( y ) = f X, Y (x, y ) dx
− −

• Conditional PDF in continuous case

• Conditional PDF of X given Y is given by
fX, Y (x, y)
fX|Y ( x | y ) = , fY ( y )  0 and
fY ( y )
• Conditional PDF of Y given X is given by

fX, Y (x, y)
fY|X ( y | x ) = , fX ( x )  0
fX ( x )
163
Probability 6.7 TERMINAL QUESTIONS
Measure and
Random 1. Joint PMF of a discrete bivariate random variable (X, Y) is shown in Table
Variable
6.14. Find joint and marginal CDF’s.

Table 6.14: Joint PMF of discrete bivariate random variable (X, Y)

Values of Y
Values of X 1 2 3 4 5 6
1 0.03 0.02 0.05 0.02 0.07 0.04
2 0.01 0.06 0.03 0.08 0.01 0.02
3 0.03 0.04 0.07 0.02 0.05 0.06
4 0.01 0.03 0.01 0.04 0.01 0.02
5 0.04 0.05 0.02 0.02 0.03 0.01

2. Some entries of the CDF of a discrete bivariate random variable (X, Y)

have lost. Available entries are given in Table 6.15 as follows.

Table 6.15: Joint CDF of discrete bivariate random variable (X, Y)

Values of Y
Values of X 1 2 3 4 5 6
1 NA 0.05 NA 0.12 0.19 NA
2 0.04 NA 0.20 NA 0.38 NA
3 NA 0.19 NA 0.46 NA 0.71
4 0.08 0.23 0.39 NA 0.69 NA
5 NA 0.32 NA 0.68 NA 1

Find the required probability in each of the following parts.

(a) ( 2  X  4, 3  Y  5 ) (b) ( 3  X  5, 2  Y  6 )
(c) (1  X  4, 2  Y  5 )

6.8 SOLUTIONS/ANSWERS
Self-Assessment Questions (SAQs)
1. (a) We know that sum of all probabilities is 1. So, we must have
2 4 2 4 2  4 4 
 p X, Y (x, y) = 1  k   (2x + 3y) = 1  k    2x +  3y  = 1
x = 0  y =1
x = 0 y =1 x = 0 y =1 y =1 
2 2
 k  ( 4(2x) + 3(1 + 2 + 3 + 4) ) = 1  k  ( 8x + 30 ) = 1
x =0 x =0

 2  2
1
 k   8x +  30  = 1  k ( 8(0 + 1 + 2) + 3(30) ) = 1  k =
 x =0 x =0  114

(b) Joint PMF of the random variable (X, Y) in formula form is given by
 (2x + 3y)
 , x = 0, 1, 2; y = 1, 2, 3, 4
p X, Y (x, y) =  114
0, otherwise
164 Joint PMF in tabular form with marginal sums is given by Table 6.16.
 Table 6.16: Joint PMF of discrete bivariate random variable (X, Y) Bivariate Random
Variable
Values of Y
Values of X 1 2 3 4 Marginal sum
3 6 9 12 30
0
114 114 114 114 114
5 8 11 14 38
1
114 114 114 114 114
7 10 13 16 46
2
114 114 114 114 114
15 24 33 42 114
Marginal sum =1
114 114 114 114 114
Using marginal sums from Table 6.16, marginal PMF of X is given in
Table 6.17.
Table 6.17: Marginal PMF of the random variable X

Values of X 0 1 2 Total
30 38 46 114
Probabilities ( p X (x) ) =1
114 114 114 114
Similarly, marginal PMF of Y is given in Table 6.18.
Table 6.18: Marginal PMF of the random variable Y

Values of Y 1 2 3 4 Total
15 24 33 42 114
Probabilities ( p Y (y) ) =1
114 114 114 114 114
(c) Joint CDF of the random variable (X, Y) is given by Table 6.19, where
probability in each cell is obtained using (6.10) repeatedly.
Table 6.19: Joint CDF of discrete bivariate random variable (X, Y)

Values of Y
Values of X 1 2 3 4
3 9 18 30
0
114 114 114 114
8 22 42 68
1
114 114 114 114
15 39 72 114
2 =1
114 114 114 114
(d) Marginal CDF of the random variable X is given by just writing first and
fifth columns of Table 6.19 together in a new Table 6.20.
Table 6.20: Marginal CDF of discrete random variable X

Values of X 0 1 2
30 68 1
FX (x)
114 114
Similarly, marginal CDF of the random variable Y is given by just writing
second and fifth rows of Table 6.19 together in a new Table 6.21.
Table 6.21: Marginal CDF of discrete random variable Y

Values of X 1 2 3 4
15 39 72 1
FX (x)
114 114 114
165
Probability (e) We can find conditional PMF of the random variable X given Y = 3 as
Measure and follows.
Random
Variable p(X = x, Y = 3)
p X|Y (x | Y = 3) = , x = 0, 1, 2
p Y (Y = 3)
  p(X = 0, Y = 3) 9 / 114 9 
9 / 33, x = 0  p X|Y (X = 0 | Y = 3) = = = 
  p Y (Y = 3) 33 / 114 33 
 
 p(X = 1, Y = 3) 11/ 114 11 
= 11/ 33, x = 1  p X|Y (X = 1| Y = 3) = = = 
  p Y (Y = 3) 33 / 114 33 

13 / 33, x = 2  p (X = 2 | Y = 3) = p(X = 2, Y = 3) = 13 / 114 = 13 
 
X|Y
p Y (Y = 3) 33 / 114 33 

(f) From Table 6.16, we have

3
( X = 0, Y = 1) = ... (6.108)
114

30 15 75
( X = 0 ) ( Y = 1) =  = ... (6.109)
114 114 114  19

From (6.108) and (6.109) it is clear that

( X = 0, Y = 1)  ( X = 0 ) ( Y = 1)
Since (6.72) does not hold so random variables X and Y are not
independent.

(g) Using (6.63), we have

(a  X  b, c  Y  d) = FX, Y (b, d) − FX, Y (a, d) − FX, Y (b, c) + FX, Y (a, c)

In our case a = 0, b = 2, c = 1, d = 3, so, we have

( 0  X  2, 1  Y  3 ) = FX, Y (2, 3) − FX, Y (0, 3) − FX, Y (2, 1) + FX, Y (0, 1)

72 18 15 3 42 Using probabilities 
= − − + =
114 114 114 114 114  from cells of Table 6.19 


2. Given CDF of the jointly continuous random variable (X, Y) is

0, if x  0 or y  0

( )(
 8x − x 10y − y
2 2
)
, 0  x  4, 0  y  5
 400

 8x − x 2
FX, Y (x, y) =  , 0  x  4, y  5 … (6.110)
 16
10y − y 2
 , x  4, 0  y  5
 25
1, if x  4 and y  5


In view of (6.107), to obtain joint PDF of the jointly continuous random

variable (X, Y), we have to differentiate (6.110) partially first w. r. t. y and
then w. r. t. x or first w. r. t. x and then w. r. t. y. So, let us first
166 differentiate (6.110) partially w. r. t y., we have
 0, if x  0 or y  0 Bivariate Random

( )
Variable
 8x − x (10 − 2y )
2

 , 0  x  4, 0  y  5
400
 
y
( FX, Y (x, y) ) = 0, 0  x  4, y  5 … (6.111)
10 − 2y
 , x  4, 0  y  5
 25
0, if x  4 and y  5


Now, differentiating (6.111) partially w. r. t. x, we get

0, if x  0 or y  0

 ( 8 − 2x )(10 − 2y ) , 0  x  4, 0  y  5
 400
2 
x  y
( X, Y
F (x, y) ) 
= 0, 0  x  4, y  5
0, x  4, 0  y  5

0, if x  4 and y  5



 ( 4 − x )( 5 − y )
2  0  x  4, 0  y  5
or
x  y
( FX, Y (x, y) ) =  100
,
0,
 otherwise

Hence, joint PDF of jointly continuous random variable (X, Y) is given by

 ( 4 − x )( 5 − y )
2  0  x  4, 0  y  5
fX, Y (x, y) =
x  y
( FX, Y (x, y) ) =  100
,
0,
 otherwise

This matches with (6.98) as expected where k = 1/100 refer to (6.99).

3. Before obtaining required probability, we have to first look at the region

formed by X + Y  2 which is shown in Fig. 6.8 (b) in blue colour. The
region of integration is triangular in shape. So, limit of only one variable
will be constant and limits of other variable will be in terms of another
variable. Such things we have discussed in detail in Unit 9 of the course
MST-011. Let us keep limits of random variable X as constant and limit
of random variable Y in terms of X. To do so, we have to draw a vertical
strip PQ (say). Strip PQ starts from y = 0 and ends at
 x y 
 Equation of line AB using intercept form is + = 1 
y=2–x 2 2
 
 You may refer to (6.18) in Unit 6 of the course MST - 011

Now, required probability is given by

2 2− x 2 2−x
(4 − x)(5 − y)
( X + Y  2) =   fX, Y (x, y) dy dx =   dy dx
0 0 0 0
100
2− x
(4 − x)  y  (4 − x)  (2 − x)2 
2 2 2
= 5y −  dx =  10 − 5x −  dx
0
100  2 0 0
100  2 

(4 − x)  20 − 10x − 4 − x 2 + 4x  (4 − x) 16 − 6x − x 2 
2 2
=   dx = 0 100   dx
0
100  2  2 
167
Probability 2
 64 − 24x − 4x 2 − 16x + 6x 2 + x 3  2
 x 3 + 2x 2 − 40x + 64 
Measure and =   dx = 0   dx
0   
200 200
Random
2
Variable 1  x 4 2x 3  1  16  43
=  + − 20x 2 + 64x  =  4+ − 80 + 128  =
200  4 3  0 200  3  150

Terminal Questions
1. Using (6.10) repeatedly on values of Table 6.14 joint CDF of the discrete
bivariate random variable (X, Y) is given by Table 6.22 as follows.
Table 6.22: Joint CDF of discrete bivariate random variable (X, Y)

Values of Y
Values of X 1 2 3 4 5 6
1 0.03 0.05 0.10 0.12 0.19 0.23
2 0.04 0.12 0.20 0.30 0.38 0.44
3 0.07 0.19 0.34 0.46 0.59 0.71
4 0.08 0.23 0.39 0.55 0.69 0.83
5 0.12 0.32 0.50 0.68 0.85 1

Marginal CDF of random variable X is given by combining values of X

from the first column and corresponding probabilities from the seventh
column of Table 6.22 and given by Table 6.23.
Table 6.23: Marginal CDF of the random variable X

Values of X 1 2 3 4 5
Probabilities ( p X (x) ) 0.23 0.44 0.71 0.83 1

Marginal CDF of random variable Y is given by combining values of Y

from the first row and corresponding probabilities from the sixth row of
Table 6.22 and given by Table 6.23.
Table 6.23: Marginal CDF of the random variable Y

Values of Y 1 2 3 4 5 6
Probabilities ( p Y (y) ) 0.12 0.32 0.50 0.68 0.85 1

2. Using (6.63), we have

(a  X  b, c  Y  d) = FX, Y (b, d) − FX, Y (a, d) − FX, Y (b, c) + FX, Y (a, c)

(a) In our case a = 2, b = 4, c = 3, d = 5, so, we have

( 2  X  4, 3  Y  5 ) = FX, Y (4, 5) − FX, Y (2, 5) − FX, Y (4, 3) + FX, Y (2, 3)

= 0.69 − 0.38 − 0.39 + 0.20 = 0.12

(b) Similarly, in this case a = 3, b = 5, c = 2, d = 6, so, we have

( 3  X  5, 2  Y  6 ) = FX, Y (5, 6) − FX, Y (3, 6) − FX, Y (5, 2) + FX, Y (3, 2)
= 1 − 0.71 − 0.32 + 0.19 = 0.16

(c) In this case a = 1, b = 4, c = 2, d = 5, so, we have

(1  X  4, 2  Y  5 ) = FX, Y (4, 5) − FX, Y (1, 5) − FX, Y (4, 2) + FX, Y (1, 2)
= 0.69 − 0.19 − 0.23 + 0.05 = 0.32
168