Lec 22 : Probability Functions for Continuous Bivariate and Multivariate Random Variables

IIT Kanpur July 201
20 followers
1 Video View·Jun 5, 2024
[Music]
Hello friends welcome to the course
multivariate procedure with r so you can
recall that in the last lecture we had
talked about the concepts of probability
density function probability Mass
functions cumulative distribution
function
Etc now in this
lecture I am going to extend them to a b
variate and multivariate
setup now if you try to recall that we
had discuss that we have two types of
random variable discrete and
continuous and uh when we are talking
about their probability function for
example probability density function for
continuous random variable and
probability Mass function for the
discrete random variable then they have
different types of mathematical
properties that we have
understood so now when we are going uh
forward and when we are trying
to uh discuss the b variate or say
multivariate random variables or random
vectors then we will have the concept of
probability density function and
probability Mass function for
this by variate random variables and
multivariate random vectors so I can
have here both the concept I can discuss
here both the
details but as I said in in a couple of
lectures that in this lecture in this
course my objective is to bring you all
to a multivariate setup and in order to
bring you to the multivar cut
up I'm trying to start from the
univarate then by variate and then
multivariate so that there is a smooth
ceiling and all these topics which I am
doing here although they are related to
the probability Theory and other topics
but they are going to help us in the
development of the tools for the
multivariate procedures in the
forthcoming lecture so I have chosen
here some selected topics which I'm
trying to explain you here so now given
the forthcoming lectures in the future I
am going to use mainly The Continuous
random variable the type of tools which
we are going to discuss about uh
multivariate analysis like as discent
analysis classification analysis Etc I'm
going to develop them for the continuous
random variable so that is why in order
to save the time I'm going to consider
only here the case of continuous random
variable and once you understand uh this
continuous case this concept can be
directly converted into the discrete
case also without much difficulty so
this is what you always have to keep in
mind when you are trying to attend this
course right so now in this lecture I
try to extend the definition of say this
probability density
function of a univariate random variable
to the setup of bivariate random
variable
T so let us begin our lecture and try to
understand it okay so as we had
discussed in the case of univariate
continuous random variable that uh we
have the probability density function
now once again we are going to
concentrate only on the continuous
random variables but now we have two
continuous random variables and we
suppose we denote them by X and Y so
definitely then now the probability
functions in such cases is going to be
affected by the
uh the behavior of X and Y both and that
is why we call it here as a join
probability distribution right and if
you try to recall that in the case of
univariate we were trying to find out
area under a curve like this on to
compute different types of probabilities
but now when you have two random
variables which are jointly affecting
the probability of an event then
essentially what is happening that we
need to calculate the probability that X
and Y assume a value in any given region
of two dimensional space right for
example now there will be suppose here
two random variables one is taking value
between a and b and Y is taking value
between say C and D then now you have to
compute the probability of occurrence of
say X and Y when X is lying between a
and b and Y is lying between C and D so
analogous to the definition of
probability density function for a
single continuous random variable we can
ALS also extend and Define the join
probability density function over the
two dimensional space right so now if
you try to recall in the earlier case we
had defined the probability density
function by these two condition FX
should be greater than zero at all point
of X and uh this integral of FXX DX
should be equal to one now I try to
extend these two condition to the setup
of a joint probability density function
for two continuous random variable X and
Y so it is now indicated here say F of
XY and so earlier it was F of XX now it
is extended to F of XY means the random
variables are denoted by uppercase
alphabet and their values are indicated
by lower case alphabets like this right
so now the first condition that F ofx
should be greater than or equal to zero
at all points can be extended to a b
variate setup as F of XY XY is greater
than equal to 0 for all values of X and
Y and the condition that minus infinity
to Infinity f of x DX should be equal to
1 that can be extended to a by variat
setup by writing two integrals minus
infinity to Infinity one integral is for
here x and one integral is for here Y
and then the function f of xyxy DX Dy is
equal to 1 and earlier we had reason we
have taken the region for X to be here R
now we have here two random variables
both are belonging to here R so this XY
will belong to say R into r or say r² so
we say here that for any region a
belonging to R2 of a two Dimension space
if you want to compute the probability
that X and Y lies in this region a this
can be obtained by integrating the
integral over the region of a with
respect to F of XY DX Dy right so
typically this F of XY is defined over
all of two dimension Space by assuming
that F of xyxy is equal to Z for all
point for Which F of XY is not a
specified right so this is simply an
extension of the the definition of the
univariate probability density function
to a viate density function so now for
example if you want to compute any
probability in the case of uh joint
probability distribution of two random
variables so now if you have a two
random variable X and you want to know
the probability that X lies between
small X1 and small X2 and capital Y lies
between small y1 and small Y2 then this
can be obtained by the integral y1 to Y2
X1 tox2 F of XY DX Dy so that is
straightforward now in case if I ask you
that you have here say p random
variables right then if you want to find
out here X1 say less than X1 say here y1
or let me write down here like this
suppose you have here the random
variables here X1 X2 here XP and suppose
these values are lying between for which
you want to compute the probability A1
B1 A2 B2 up to here AP BP and so on then
what will be this will be the
integration over see here X1 X2 here XP
like here A1 to B1 A2 to a22 B2 up to
here ap2 BP F of X1 X2 XP values are
here X1 X2 XP dx1 dx2 dxp like this so
this is my objective this is what I want
to show you right now when you are
trying to have the concept of join
probability density function then there
is an Associated concept of marginal
probability distribution what is the
difference between the two uh suppose
you have two random variables which are
going to affect the probability of an
event and their behavior is actually
joint that both of them are going to
affect the probability of an event now
you want to know that what is the
behavior of individual random variables
it is something like if I take a very
crude example that suppose in a class A
group of students are assigned a project
that means uh the class is divided into
different groups and every group has
suppose five students and now every
student is doing a group project so when
the project outcome comes then it is
very difficult to decide that what is
the contribution of an individual
student but we see only the final
outcome so the final outcome is
something like joint probability that it
is the effect of all the five student
but if I want to find out the the effect
of only one student then that will be
equivalent to Marginal distribution or
say marginal contribution or
equivalently marginal probability
distribution right so if I consider here
two random variable X and Y so so I can
find out here a marginal probability
distribution from The Joint probability
distribution function and exactly in the
same way as we try to find out the
probability function if it is continuous
then integral and if it is discrete then
here is sum so an analogous approach is
used to determine the marginal
probability distribution from The Joint
probability density function what
exactly you want to do you want to
remove suppose if I want to find out the
marginal of here
x like this so what essentially I have
to do I have to remove the effect of
here y from here so now how to get it
done in order to remove the effect
mathematically how what you can think of
you see whenever we want to do something
we have to think of a mathematical
procedure that can executed so in case
if I try to integrate out the effect of
y y is effect then the remaining part
will be treated as a effect of only X so
that is what we try to do in the case of
marginal
probability distribution so if there are
more than one random variables that are
associated with the random experiment we
distinguish between the joint
probability distribution of X and Y and
the probability distribution of each
variable individually that mean the
probability distribution of X
probability distribution of Y and we
have a joint probability distribution of
X and Y and this individual probability
distribution of a random variable they
are referred to as marginal probability
distribution for example marginal
probability distribution of X marginal
probability distribution of Y right so
in general the marginal probability
distribution of X can be determined from
The Joint probability distribution of X
and other random variables in general
right so let me try to show you here in
the case of bivariate which can be very
easily extended to a multivariate case
so if I have here A B variate
probability density function fxy f FY
now you want to find out here marginal
distribution of X which is indicated
here by f of x so now I'm trying to
integrate out the effect of Y just by
integrating this joint probability
density function over the range of Y and
so it will be the range of here Y and
here Dy right
so now I believe that the effect of Y
has been taken out and now whatever is
the effect remaining that is that
belongs to X and similarly if you want
to find out the marginal distribution of
Y then it is something like you try to
take here the joint rective functional
function and try to integrate it out
over the range of here x like this
integral over X fxy XY DX and now if I
ask you suppose if you want to extend it
to a multivariate setup suppose if you
have here three variables x y z and
based on that you have
here a probability density function need
a join density function f of XY Z and
suppose if you want to find out here the
effect of here x f of x so what it now
will become so f of x x will become here
integral over y integral or z f of x x y
z XY Z now this will be over Dy and DZ
and similarly if I try to take here more
than three variables suppose if I have
random variable X1 X2 X3 X4 X5 and
suppose I want to find out here uh The
Joint effect of X1 X2 by removing the
effect of x3 X4
X5 so for that obvious viously you will
have here a probability quantity
function f of X1 X2 X3 X4
X5 like
this and now you want to obtain here
suppose F of X1
X2 so what are you going to do here you
try to write down this function and try
to integrate it over say X3 X4 X5 dx3
dx4
dx5 so this is how you can extend this
univariate concept to a multivariate
concept right and similarly we also have
here concept of conditional density
function or conditional distribution so
when we have a joint density function
say FX f ofx y then if I want to find
out the individual contribution of X and
Y then we use the concept of marginal
distribution but suppose if I want to
see the effect of one variable when the
effect of another variable is fixed then
how to find out uh this effect so this
is obtained by the concept of
conditional probability distribution so
if I have here two random variables X
and Y then the conditional distribution
of x given capital y equal to Y is
obtained here like this so what is
happen
the value of y here is fixed now I want
to see the behavior of X that how the
probabilities are changing so this is
indicated by here the random variable
like as here x given Y is equal to y or
it is WR as a capital x this vertical
line and then here is small y so now you
are you want to find out the conditional
density of X so it is written here the
random variable here is x given Y and
then here this is here X so this will be
here joint probability density function
divided by the marginal density function
so here in the
numerator this is
Joint
probability
density of X and Y and in the
denominator this is the
marginal density of of Y so this is how
you can obtain
the distribution of the conditional
probabilities and similarly if you want
to change the role of X and Y then the
conditional distribution of Y given X is
here like this you can see here now here
the random variable is y given X and the
values which are given is here Y and it
is here the
joint PDF in the numerator and
marginal
density of X in the
denominator right this concept can be
extended to a multivariate setup
also right
suppose if I say there are here suppose
here four random variables X1 X2 X3 X4
and suppose I want to see The Joint
effect of X1 X2 condition that
X3 and X4 they are here
fixed so similar type of definition can
be given here also in the numerator this
will be X1 X2 X3
X4 and in the it is going to be here F
of see here
X3
X4 X3 X4 like this right so now you can
see here this Concepts can be extended
to a multivariate setup without any
problem now let me try to take some
example to explain you how this
probabilities can be computed they are
very simple example just to give you an
illustration suppose we have a joint
probability density function of X and Y
which is given here by this function
which takes Val tce of exponential of -
x into exponential of - 2 y when X lies
between 0 and infinity and Y Li between
Z and infinity and zero otherwise
elsewhere suppose if you want to compute
the abity that X greater than 1 and y
less than 1 so this type of integral
with the range of X between 0 and 1 and
that means range for y as 1 to Infinity
this joint PDF this can be solved I'm
not going to solve here this integral
but anyway you can obtain this value
similarly if you want to obtain here the
probability of X less than y now it is a
joint variation so you can write down
here the limits of X as 0 to infinity
and probability of and the limits for y
as 0 to Y write down the PDF and solve
it this will come out to be here 1 upon
3 and similarly if you want to just find
out that probability of X is less than C
that mean probability and that mean the
Y is lying between Z to Infinity you can
have this integral range for x and 0 to
Infinity range for y The Joint PDF and
then you can solve this integral without
any problem right so now you can see
here different types of these
probabilities can be computed from this
joint probility density function now we
consider another example to show you the
application of marginal distribution
suppose our joint PDF is given by here
like this that F of XY is equal to 2 if
0 is less than x less than y less than 1
and zero otherwise if you want to find
out here the marginal of here y then it
is here you have to integrate over the
range of X which is here 0 to Y and then
this joint PDF and if you solve this it
will come out to be here tce of Y and
similarly if you want to find out here
the marginal of here x then you have to
integrate over the range of here y over
the PDF and this will come here like so
you can see here it is only an an
integration integration problem that you
have learned in your undergraduate and
similarly if you want to find out the
conditional dity function in such a case
so now you already have found here this
marginal density so by the definition of
conditional density function of x given
y this is here joint PDF divided by
marginal PDF of Y so it will come out to
be 2 upon 2 Y which is here like this 1
upon Y for the range X line between 0
and Y and zero otherwise and if you
similarly if you want to find out the
conditional density of Y given X this is
the joint PDF of X and Y divided by the
marginal of here x which is equal to
here like this you already have obtained
this marginal PDF and the joint PDF here
is given by two and this is the marginal
PDF so it will be here like this
so you can see here finding out this
conditional and marginals are equivalent
to finding out any integral problem that
solve right now after this we have
another concept of stochastic
Independence which is in usual word we
we call it as independent random
variables or Independence so what is the
basic concept of Independence that if
you have more than one random variables
suppose I have here two random variables
X and Y and suppose if the occurrence of
x does not affect the probability of Y
and if the occurrence of Y does not
affect the means occurrence of X or the
probability of X that means X and Y they
are not affecting to each other then we
say that they are independent right so
two continuous random variable X and Y
are said to be stochastically
independent stochastic means random if
the joint density function of X and Y
can be expressed as the product of their
marginal density function so the joint
density function of X and Y can be EXP
as the marginal density function of X
and marginal density function of Y and
the same concept can be extended to
discrete random variable also that uh
The Joint probability that xal to Small
x i and y equal to small YJ this can be
expressed as the product of their
individual probabilities that
probability xal to x i into probability
that y = to YJ for all value of
I and the same condition for independent
in terms of a cumulative distribution
function can also be expressed so in
terms of joint CDF for example here like
this F of
XY the random variable X and Y are
independent if the joint CDF can be
expressed as the product of their
product of the CDF of individual
variable that is f of XY can be
expressed as the product of f of x and F
of Y right so if you want to see through
an example I can consider here joint
probability distribution function here
like this and I try to find out here the
marginal density of here x which is here
integral over the range of Y and then
the same integral now you can solve this
integral it will come out to here like
this right similarly if you try to find
out the marginal density of Y it will be
integrated over the range of here x into
with this probability density function
and if you try to solve this integral
this will come out to be here like this
so now you can see here that the product
of this and this this is the same as
this given joint density function that
is f of XY is equal to F ofx into F of Y
and this implies that X and Y are
independent so this is how we try to try
to find out that if any set of random
variable they are independent or not now
the same concept can be extended to a
multivariate setup also if I see here if
I have got here P random variables then
then if they can be expressed as F of X1
F of X2 up to here F of XP and the same
concept is true for the CDF
also right up to here F of
XP right so now you can see here how I
have extended the univariate and
bivariate concept to a multivariate
setup but definitely I will try to
introduce you these Concept in a multiv
setup when we try to consider the multiv
normal distribution but here I would
like to conclude it by
defining the way we are going to
consider the random
variables uh or a random Vector so
suppose we have P random variables which
are indicated by X1 X2 XP and now they
can be combined in a vector which is a
matrix type Vector it's not a data
Vector because what I'm trying to say
say so all these variable can be
combined in a random Vector of order say
p cross1 like this xcore and these are
here this random variables X1 X2 XP and
they can be also expressed here as X1 X2
XP transpose and when you are trying to
find out the space of it this is going
to be the set of N duples and now you
can Define here PDF something like here
X1 X2 XP X1 X2 here XP or in brief it
can be written here let's say f of x and
like here at as this and then based on
this you can Define different types of
marginal density function conditional
density function Etc right so in general
we will try to denote the PDF of the
random Vector X by here like this and
similarly we can Define the marginal
distribution Independence of the random
vectors can be decided but anyway we
will try to handle it when we try to
consider a multiv normal density
function right so so now we come to an
end to this lecture so now you can see
what we have done in this lecture
earlier lecture we introduced the
univariate random variable now in this
lecture I extended the same concept to a
bivariate random variable and very
silently and gradually I extended the
same concept to a multivariate setup so
we had univariate random VAR variable by
variate random variable and now we have
a random Vector the concept of marginal
distributions the concept of conditional
distribution they occur only when we
have at least two random variables
because you need something to
marginalize or something to be
conditioned but the concept of marginal
distribution and conditional
distribution that I have explained in
the context of B variate random variable
they can be extended to the multivariate
setup also in the B various setup I had
only one variable one variable for the
conditional but in the multivar setup I
can have two random sub vectors when we
are trying to talk about the
independence we can talk about the
independence of the individual
components we can talk about the
independence of the random sub vectors
also for example if I have five random
variabl X1 X2 X3 X4 X5 support X1 and X2
and another set of variable X3 X4 X5
they are
independent so all these Concepts they
are derived from this very simple
concept but they are extended to a
situation and they are modified
according to the need of the time so my
request is that you try to pick up any
book and try to at least choose some
example on the by
variate probability density function and
try to uh just see how you can Define
them for discrete random variable how
you can Define them for the continuous
random variable and how you can obtain
different types of marginal density
function conditional density function
and definitely this will make you more
confident once you are confident then
nobody can stop you in excelling in the
life so you try to practice it and I
will see you in the next lecture with
more topics on this probability
functions till then
[Music]
goodbye
[Music]
[Music]
