Introduction to Probability Theory and Stochastic Processes

Prof. S. Dharmaraja
Department of Mathematics
Indian Institute of Technology, Delhi

Lecture - 27

In this lecture we are going to discuss the independent random variables. When we
discuss many random variables, sometimes some random variable they may have a
relations within themselves or sometimes may not. So, it is important to study whether
these random variables are having some dependency or not. So, this dependency can be
studied by using a nice mathematical way through the CDF that is a joint CDF. If the
joint CDF of two dimensional random variable or n dimensional random variable satisfy
some conditions then we can conclude those random variables are independent.

Then the next question comes why do you need to study the independent of random
variables. If the random variables are independent, then some of the prediction or some
of the sum of finding the probabilities of those random variables will be easy when those
random variables are independent. So, let me start with the concept called independent
random variables.

(Refer Slide Time: 01:21)

Let me start with the definition of independent random variables, then few more
properties when this random variables are independent at the end I will give one or 2
examples for the conceptual understanding the definition.

(Refer Slide Time: 01:54)

Let me start with the two dimensional random variable, then the same concept can be
extended to the n dimensional random variable. So, that way it is easy to explain the
concept. We say that the random variable X and the random variable Y or independent if
and only if the CDF of two dimensional random variable that is same as the CDF of one
dimensional random variable with the product. Whenever I write the suffix; that means,
the CDF is corresponding to that random variables. So, F suffix x comma y; that means,
it is the CDF of two dimensional random variable F suffix x.

That means x is the CDF of the random variable x, F suffix capital Y; that means, CDF of
the random variable y. If the product of CDFs of the random variable x and y that is same
as the CDF of two dimensional random variable x comma y for all x comma y in R 2.
Then we can conclude this is a independent random variables they say if and only if
condition; that means, if two random variables are independent this condition will be
satisfied, if this condition is satisfied then we can conclude both the random variables are
independent. It is a immaterial of the random variables are of the discrete type or
continuous type or mixed type because the CDF is always exist whatever be the type of
random variable. Here we are saying the two random variables are independent if and
only if the condition on the CDFs.

That means, if I have a for any pair of Boral sets. Suppose I keep the Boral set A and
another Boral set B of real line then we have if two random variables are independent,
then the probability of x belonging to the Boral set A and y belonging to the Boral set B
that is same as the probability of x belonging to the Boral set A multiplied by the
probability of y belonging to Boral set B. That means, if two random variables are
independent if and only if conditions is satisfied.

Whenever two random variables are independent, then we can always get the probability
of x belonging to Boral set and y belonging to another Boral set that is same as X
belonging to the one Boral set multiplied by probability of Y belonging to the other Boral
set. This is if this condition is satisfied then X and Y or set b independent whereas, the
first condition is if and only if condition we are not saying the random variable is a
discrete type or continuous type. Now, I am going to make a condition on whether the
random variables are of the discrete type or continuous type and how this if and only if
condition changes that is a necessary.

(Refer Slide Time: 06:27)

And sufficient condition for that random variables x and y of the discrete type to be
independent is the joint probability mass function of x and y.
That is same as the product of the probability mass function of x with the probability
mass function of y this is for all pairs xi comma yj. That means, this is a if and only if
condition if both the random variables are of the discrete type, then the c condition
independent condition of CDF can be replaced by the independent condition on the joint
probability mass function or joint probability mass function is same as product of
probability mass functions of x and y. This is also if and only if condition; that means, if
two random variables are of the discrete type or independent then this condition will be
satisfied we call this condition as a independent condition.

Similarly, if this condition is satisfied for all pairs, then both the random variables are of
the discrete type and they are independent.

(Refer Slide Time: 08:40)

The similar results is for the continuous type also. two random variables x and y of the
continuous type or independent, if and only if and only if you have to replace the
condition of the CDF by the probability density function. So, the joint probability density
function when I write suffix x comma y; that means, it is a joint probability density
function of x comma y, that is same as a the probability density function of x multiplied
by probability density function of y this is for all x comma y. So, if this condition is
satisfied then two random variables are of the continuous type or independent.

If two random variables are of the continuous type or independent, then this condition
will be satisfied for all x comma y. Therefore, the condition for independent random
variables either in the level of a CDF or if it is a discrete random variable in the form of
probability mass function if the random variables are of the continuous type, then it is a
probability density function for. So, all are all 3 are going to be if and only if condition
not the only one side it is in the both side. So, even though we have explained through
the 1 d two dimensional random variables this can be extended to the multi dimensional
random variable also.

(Refer Slide Time: 10:49)

That means, let x 1 comma x 2 comma so, on be a n dimensional random variables with
the CDF capital F is a function of n variables. We say that the random variable x 1 x 2 xn
are mutually independent are mutually independent random variables, if and only if you
take any fewer random variables CDF k, for K is equal to 2 3 and so, on till n, that is
same as product of i is equal to 1 to k the CDF of those random variables CDF. This
means if you take any two random variables the CDF of those two random variable is
same as a product of a CDF of only those two random variables.

If I take any 5 random variables, when n is greater than 5 then CDF of 5 random
variables is same as product of those 5 random variables CDF, then we conclude they are
mutually independent. It is same as the mutually independent it is same as the mutually
independent events. If you have a n events and once you say that they are pairwise
dependent. That means, any 2 ra events satisfies the independent concept or independent
condition then they are called pairwise. If it means mutually if they are mutually
independent; that means, whatever be the collection of events you take the independent
condition is satisfied then we conclude their mutually independent events.

The same thing here if you have n random variables whatever be the number of random
variable you take it from those n random variables, that satisfies independent condition
then it start from any 2 till all the random variable then we conclude they are mutually
independent. Whenever we say more than two random variables are independent; that
means, by default they are mutually independent whenever we say more than two
random variables are independent; that means, by default they are mutually independent.

Many times we would not write again and again mutually independent word when we
discuss more than two random variables, but whenever we have more than two random
variables when we use the word independent random variable. That means, they are
mutually dependent; that means, the independent condition is satisfied for all the forms
of collection of random variables, satisfying the independent condition.

Now, we will move into some problem of explaining how the independent random
variable is playing a role.

(Refer Slide Time: 14:41)

The first example that is same as the example which we have considered earlier, that is
let x comma y be a continuous type random variable two dimensional continuous type
random variable with joint probability density function is of the form f of x comma y,
that is after we find the value of k we got a 24 x square divided by y cube, when x is lies
between 0 to 1 and y is greater than 2 otherwise it is 0. So, this problem just now we
have discussed when two random variables are of the continuous type for the 2 we found
the k, that k value was 24 and we found the probability of x between some interval.

Here, we will verify whether these two random variables are independent or not. So, this
is the joint probability density function already in the same example we got the
probability density function of x, that is 3 times x square when x is lies between 0 to 2 1
otherwise 0. Similarly we got the probability density function of y that is 8 divided by y
cube when y is greater than 2 0 otherwise.

Easily it can be verified in this example the joint probability density function is same as
product of probability function of random variables x comma y. Because it is 3 times x
square the other one is 8 divided by y cube, not only that the 3 x square is a range
between 0 to 1 and 8 divided by y cube the range is great y is greater than 2, which is
same as if you make a product that is 24 x square divided by y cube and the range of x is
0 to one and the range of y is 2 to infinity that is same as the joint probability density
function of a 24 x square divided by y cube when x lies between 0 to 1 and y is greater
than 2.

So, their interval matches and the value matches 0 otherwise matches therefore, for all x
comma y the joint probability density function is immersed the product of density
functions of x and y therefore, these two random variables independent. Therefore, x and
y are independent random variables. Each one is of the continuous type, we can check it
from the CDF also that is if and only if condition for the CDF. But, since the joint
probability density function is given you can find the probability density function of x
and y, then you can verify the independent conditions on probability density function that
is satisfied. Therefore, both the random variables are independent.
(Refer Slide Time: 18:40)

We will go for one more example, example 2 again we have a continuous type let x and y
be a random variables of continuous type with joint probability density function that is f
of x comma y, which takes a value 6 when x is lies between 0 to 1 and y is lies between 0
to 1 as well as 3 y is less than x less than or equal to x. So, the joint probability density
function is greater than 0, that is 6 when x is lies between 0 to 1, y is lies between 0 to 1
and 3 y is less than or equal to x 0 otherwise.

Before we proceed the problem we can always verify whether this is correct joint
probability density function; that means, if you integrate if you integrate double
integration over x and y this has to be 1. So, one can verify this is going to be double
integration is one. Therefore, this is the correct joint probability density function. From
these one can find the marginal distribution of x by integrating a the joint probability
density function with respect to y. If you do the little simplification you can get the
answer that is 2 times x when x is lies between 0 to one otherwise it is 0.

You can verify this result also whether this is a correct probability density function of x
by integrating 0 to 1 to x you will get the value 1 and it is greater than or equal to 0
therefore, this is the probability density function of x. Similarly you can compute the
probability density function of y by integrating the joint probability density function with
respect to x. Here also am skipping the integration part one can get the answer that is a 6
times 1 minus 3 y, when y is lies between 0 to one third 0 otherwise; that means, within
this interval 0 to one third the probability density function is greater than 0, 6 times 1
minus 3 y otherwise 0.

By seeing the probability density function of x probability density function of y this

product is not going to be the joint; that means, the f of x comma y is not equal to the
product of probability density function of x and y. For all x comma y if this condition is
satisfied equal to then you can conclude they are independent.

But since by seeing this you can say it is a 2 times x 6 times one minus 3 y whereas, the
nine probability density function is 6 obviously you can say they are not equal.
Therefore, x and y are not independent random variables. So, I have given the first
example which in which they are independent random variable by finding the marginal
distribution whereas, in this example by finding the marginal distribution of x and y, we
are concluding condition is not satisfied independent condition is not satisfied therefore
they are not independent random variables.

(Refer Slide Time: 23:27)

We will go to the one more example that is example 3, because already we discussed 2
problems of the continuous type we will see one problem of the discrete type also, that is
let x and y be a discrete type.

Let x and y be discrete type random variables with joint probability mass function is
given by. If you recall this is same as the problem which we have discussed in the last
class, the possible values of x is 0 1 2 3 and the possible values of y is 1 and 3 where x
denotes a number of heads obtained when we tossing a unbiased coin 3 times. And y is
the difference in absolute of a number of heads and number of tails of time therefore, the
possible values of y is 1 and 3 and the possible values of x is 0 1 2 and 3, and in that
problem we have got the joint probability mass function that is 0 3 by 8, 3 by 8, 0 then 1
by 8, 0 0 and 1 by 8 and in that problem we have got the marginal distribution of x and y
also if you recall.

So, for possible values of x that is 0 1 2 and 3 and the probability of x takes a value of x
that is going to be 1 by 8, 3 by 8 again 3 by 8 and 1 by 8. So, this is a probability mass
function of x and similarly probability mass function of y that is 1 and 3. So, 1 it takes a
value 6 by 8, and 3 takes a value 2 by 8. So, this is the probability mass function of y.
Now you can verify whether this two random variables are independent suppose x takes
a value 1 y takes a value one that probabilities 3 by 8.

Here x takes a value 1 and y takes a value 1 that is this much. So, if you make product of
3 by 8 into 6 by 8 that is not equal to 3 by 8. Even at one pair it is not satisfied then you
cannot conclude it is independent random variable. If all the pairs the joint probability
mass function is same as product of probability mass functions of x and y, then only you
can conclude their independent. Since any one pair does not satisfy then you can
immediately conclude both the random variables are not independent.

This will be a set of obvious because the random variable y is defined as difference in
absolute with the number of heads and number of tails whereas, the random variable x is
defined number of heads; that means, the y itself is a function of x; that means, y is at
dependent on x. Therefore, there is a dependency between the random variable y and x in
the definition itself. Therefore, from there itself you can conclude they are not a
independent random variable, but that we have concluded from the distributions also.