0% found this document useful (0 votes)
0 views70 pages

Distribution of Functions of Random Variables

The document discusses various techniques for finding the distribution of functions of several random variables, including the distribution function technique, change of variable technique, and generating function technique. It provides detailed examples illustrating how to derive the probability density function (p.d.f) of new random variables defined as functions of existing random variables. Additionally, it explains the process of obtaining joint probability density functions through transformations and the necessary conditions for these transformations to be valid.

Uploaded by

washingtoneoburu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
0 views70 pages

Distribution of Functions of Random Variables

The document discusses various techniques for finding the distribution of functions of several random variables, including the distribution function technique, change of variable technique, and generating function technique. It provides detailed examples illustrating how to derive the probability density function (p.d.f) of new random variables defined as functions of existing random variables. Additionally, it explains the process of obtaining joint probability density functions through transformations and the necessary conditions for these transformations to be valid.

Uploaded by

washingtoneoburu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 70

DISTRIBUTION OF FUNCTIONS OF

several RANDOM VARIABLES


Introduction

There are several different techniques for finding the distribution


of functions of random variables, among them including:
(i) the distribution function technique,
(ii) the change of variable technique
(iii) the generating function technique.
Distribution function technique
 This technique is mainly used to obtain the probability density function of a newly
defined random variable which is a function of several other random variables.

 This technique is carried out as follows:


(1) First obtaining the cumulative distribution function of the new
variable using the j.p.d.f of the several random variables

(2) differentiating the cumulative distribution to obtain the required


probability density function.
Let X ,Y be jointly distributed with j.p.d.f f ( x, y )

Define a new random variable Z such that Z h ( x , y )

Then G ( z ) Pr( Z  z ) Pr( h( x, y )  z )


(1)

 

f ( x, y )dxdy
h ( x , y ) z

d
(2) g ( z)  G( z)
dy
Example 1

Let X , Y have j.p.d.f

Find the p.d.f of Z  X  Y


The range of values X , Y such that x  y  z is either

(i) 0  x z  y (ii) 0  x z  y
0  y z 0  y z
Solving the inner integral
Next solving the second integral
Thus

To get the p.d.f g(z) :


The p.d.f of Z is
Let X 1 , X 2 ,..., X k be jointly distributed with j.p.d.f f ( x1 , x2 ,..., xk )

Define a new random variable Y such that Y  g ( x1 , x2 ,..., xk )

Then
(1) F ( y ) Pr(Y  y ) Pr( g ( x1 , x2 ,.., xk )  y )

 
..
 
f ( x1 , x2 ,.., xk )dx1dx2 ...dxk
g ( x1 , x2 ,.., xk ) y

d
(2) f ( y)  F ( y)
dy
Example 2


Let X  X 1 X 2 X 3  have j.p.d.f

4e  (2 x1  2 x2  x3 ) x1  0, x2  0, x3  0
f ( x1 , x2 , x3 ) 
 0, elsewhere

Find the p.d.f of Y  X 1  X 2  X 3


F ( y ) Pr(Y  y ) Pr( X 1  X 2  X 3  y )

 

4e  (2 x1  2 x2  x3 ) dx1dx2 dx3
x1  x2  x3 y

The range of values for the three variables X 1 , X 2 , X 3 such that


x1  x2  x3  y is either

(i) 0  x3  y  x1  x2 (ii) 0  x3  y  x1  x2
0  x2  y  x1 0  x1  y  x2
0  x1  y 0  x2  y
(iii) 0  x1  y  x3  x2 (iv) 0  x1  y  x3  x2
0  x2  y  x3 0  x3  y  x2
0  x3  y 0  x2  y

0  x2  y  x3  x1 0  x2  y  x3  x1
(v) (vi)
0  x1  y  x3 0  x3  y  x1
0  x3  y 0  x1  y
y y  x3 y  x2  x3

F ( y )   4e (2 x1  2 x2  x3 ) dx1 dx2 dx3


0 0 0

Solving the innermost integral


y  x2  x3 y  x2  x3


0
4e  (2 x1  2 x2  x3 ) dx1 e  (2 x2  x3 ) 
0
4e  2 x1 dx1

 (2 x2  x3 )  2 x1 y  x2  x3
e   2e 
0

e  (2 x2  x3 )  2  2e  2( y  x2  x3 ) 
2e  (2 x2  x3 )  2e  (2 y  x3 )
Solving the innermost integral
y  x2  x3 y  x 2  x3

 
 (2 x1  2 x2  x3 )  (2 x2  x3 )  2 x1
4e dx1 e 4e dx1
0 0

 (2 x2  x3 )  2 x1 y  x2  x3
e   2e 
0
 (2 x2  x3 )
e  2  2e 2( y  x2  x3 ) 
 (2 x2  x3 )  (2 y  x3 )
 2e  2e
Next solving the second integral

y  x3 y  x3 y  x3

  
 (2 x2  x3 )  (2 y  x3 )  x3  2 x2  (2 y  x3 )
2e  2e dx 2 e 2e dx 2  2e dx 2
0 0 0

 x3  2 x2 y  x3  (2 y  x3 ) y  x3
e   e    2 x2 e 
0 0

e  x3 (1  e  2( y  x3 ) )  2( y  x3 )e  (2 y  x3 )
Finally solving the third integral
y


 x3  2( y  x3 )  (2 y  x3 )
e (1  e )  2( y  x3 ) e dx3
0
y y y y

e  x3 dx3  e  2 y e x3 dx3  2 ye 2 y  3


e x3
dx  2e  2y
 3 dx3
x e x3

0 0 0 0

 x3 y y y y
  e   e  2y
 e   2 ye
x3  2y
 e   2e
x3  2y
 x3 e  e 
x3 x3
0 0 0 0

1  e  y  e  y  e 2 y  2 ye y  2 ye 2 y  2 ye y  2e y  2e  y

1  2e  y  e  2 y  2 ye 2 y
Thus
F ( y ) 1  2e  y  e  2 y  2 ye  2 y

To get the p.d.f f(y) :

d d
f ( y )  F ( y )   1  2e  y  e 2 y  2 ye 2 y 
dy dy
2e  y  2e 2 y  4 ye 2 y  2e 2 y
2e  y  4 ye 2 y
The p.d.f of Y is

 2e  y  4 ye  2 y y 0
f ( y ) 
0, elsewhere
Example 3

Let X 1 , X 2 ,..., X n be independent random variables , each with p.d.f

 3(1  x) 2 0  x 1
f ( x ) 
0, elsewhere

Find the p.d.f of Y=min(X1 , X2 , …., Xn )


F ( y ) Pr  min( X 1 , X 2 ,..., X n )  y 

 
2 n
 ...... [3(1  xi ) ] dx1dx2 ...dxn
   
min( x1 , x2 ,.., xn ) y

This is equivalent to
F ( y ) Pr  min( X 1 , X 2 ,..., X n )  y 
1  Pr  min( X 1 , X 2 ,..., X n )  y 
1  Pr  X  y1 , X 2  y,..., X n  y 

n
1  Pr( X
i 1
i  y)
1
3 1
Pr( X i  y ) 3(1  xi ) dxi    (1  xi ) 
2
y
y

0  ( (1  y )3 ) (1  y )3

n n

 i
Pr(
i 1
X  y )  (1  y )
i 1
3
 (1  y ) 3n

Thus F ( y ) 1  (1  y )3n
d d
f ( y )  F ( y )  1  (1  y ) 3n 3n(1  y ) 3n  1
dy dy

The p.d.f of Y is

3n(1  y )3n  1 0  y 1
f ( y ) 
 0, elsewhere
Change of variable technique

Let X , Y be jointly distributed with j.p.d.f f ( x, y )

Define two new random variables U , W such that each of them is


a function of the variables X , Y i.e.

U  g1 ( X , Y )
W g2 ( X ,Y )
If
(i) the functions
X h1 (U ,W )
Y h2 (U ,W )

can be determined so that there is a one-to-one correspondence ;


(ii) The partial differentials

can be derived; then the j.pd.f of U , W can be obtained and it is given by


 
g u, w   f  h1 (u , w) , h2 (u , w)  J
  x    y   
  absolute value of J

where J is the determinant of the matrix of partial derivatives

 x x 
 u w 
 
 y y 
 u w 

it is referred to as the Jacobian of transformation


Example 4

Let X and Y be two random variables with j.p.d.f

f ( x, y ) e ( x  y ) x  0, y  0

Find the j.p.d.f of two variables U and W such that

U=X-Y and W=X+Y


from U X  Y ; W X Y

U W UW
we have X  and Y 
2 2

The range of values for U and W are:


uw
0x <  0 < u+w < 
2

w u
0 y    0  w u  
2
u  w and u   w   w  u  w ; w  0
The partial derivatives required to obtain the Jacobian of transformation are:

x  (u  w) 1 x  (u  w) 1
   
u u 2 2 w w 2 2

y  ( w  u ) 1 y  (w  u ) 1
   
u u 2 2 w u 2 2

 1 1
 2 2   1 1    1 1 1
J det      2 2   2
 1 1   2 2   
 2 2 
The j.p.d.f of U and W is given by

 
g u, w   f  h1 (u , w) , h2 (u , w)  J
  x    y   
  absolute value of J

u w w u 1 1 w
exp      e
 2 2  2 2
The j.p.d.f of U and W is

1 w
g u , w   e  w  u  w; w  0
2
Let X 1 , X 2 ,..., X k be jointly distributed with j.p.d.f f ( x1 , x2 ,..., xk )

Define a new random vector Y Y1 Y2 ....Yk  such that each random
variable in the newly defined random vector is a function of the variables
X 1 , X 2 ,..., X k i.e.

Y j  g j ( X 1 , X 2 ,..., X k )
If
(i) The functions Y j  g j ( X 1 , X 2 ,..., X k ) are partially
differentiable with respect to each variable Xj

X j h j (Y1 , Y2 ,..., Yk )
(ii) the functions can be determined so that
there is a one-to-one correspondence ;

Then the j.pd.f of Y can be obtained and it is given by


 
g  y1 , y2 ,..., yk   f  g1 ( y1 ,.., y k ) , g 2 ( y1 ,.., y k ) ,...., g k ( y1 ,.., y k )  J
               
 x1 x2 xk 

Where J is the Jacobian of transformation


Example 3

Let X  X 1 X 2 X 3  have j.p.d.f

e  ( x1  x2  x3 ) x1  0, x2  0, x3  0
f ( x1 , x2 , x3 ) 
 0, elsewhere
Find the p.d.f of Y Y1 Y2 Y3  such that

X1 X2
Y1  ; Y2  ; Y3  X 1  X 2  X 3
X1  X 2  X 3 X1  X 2  X 3
We begin by obtaining the functions of X 1 , X 2 , X 3 in terms of
Y1 , Y2 , Y3
Given that Y3  X 1  X 2  X 3 ;

X1 X1
Y1    X1 Y1Y3
X1  X 2  X 3 Y3
X2 X1
Y2    X 2 Y2Y3
X1  X 2  X 3 Y3
X 3 Y3  X 1  X 2 Y3  Y1Y3  Y2Y3
 x1 x1 x1   y1 y3 y1 y3 y1 y 3 
   
 y1 y2 y3 
 y1 y 2 y 3

 x2 x2 x2   y2 y3 y 2 y3 y 2 y 3 
J    
 y1 y2 y3 

y1 y 2 y 3

 x3 x3 x3   y3  y3 y1  y3 y 2 y3  y3 y1  y 3 y 2 y 3  y 3 y1  y3 y 2 
   y1 y 2 y 3 
 y1 y2 y3 

 y3 0 y1 
  0 y3 y2 

  y3  y3 1  y1  y 2 
 y3 0 y1 
 0 
 y3 y2   y3  y3 (1  y1  y2 )  ( y2 y3 )  
  y3  y3 1  y1  y2 
y1  (0  y3 )  ( y3 ) 
 2

 y3  y3  y3 y1  y3 y2  y2 y3   y1 y32  y32  y32 y1  y1 y32  y32


g ( y1 , y2 , y3 ) e  ( y1 y3  y2 y3  y3  y1 y3  y2 y3 ) y32  y32 e  y3

We need to determine the range of values for the new variables

X1
Y1   0  y1  1
X1  X 2  X 3
X2
Y2   0  y2  1
X1  X 2  X 3
Y3  X 1  X 2  X 3  y3  0
 y32 e  y3 0  y1  1;0  y1  1; y3  0
g ( y1 , y2 , y3 ) 
0, elsewhere
Example 4
Let X 1 , X 2 , X 3 be independent random variables, each with p.d.f

f ( xi )  xi e  xi xi  0

Find the j.p.d.f of Y1 , Y2 , Y3 such that

Y1  X 1  X 2  X 3 ; Y2  X 2  X 3 ; Y3  X 3
Defining X 1 , X 2 , X 3 in terms of Y1 , Y2 , Y3

Y3  X 3  X 3 Y3
Y2  X 2  X 3  X 2  Y3  X 2 Y2  Y3
Y1  X 1  X 2  X 3  X 1  Y2  X 1 Y1  Y2
 x1 x1 x1   ( y1  y2 ) ( y1  y2 ) ( y1  y2 ) 
   
 y1 y 2 y3
  y1 y 2 y3


J 
x2 x2 x2  

 ( y 2  y3 ) ( y 2  y3 ) ( y 2  y 3 ) 
y1 y 2 y3  y1 y 2 y 3 
   
 x3 x3 x3   y3 y 3 y 3 
 y1 y 2 y3   y1 y 2 y 3 

1  1 0 
 
  0 1  1 1
 0 0 1 
g ( y1 , y2 , y3 ) ( y1  y2 )( y2  y3 ) y3e  ( y1  y2  y2  y3  y3 ) 1

 ( y1  y2 )( y2  y3 ) y3e  y1

We need to determine the range of values for the new variables

Y1  X 1  X 2  X 3  y1  0
Y2  X 2  X 3  y2  0
Y3  X 3  y3  0
Also x2  0  y2  y3  0, y3  y 2
x1  0  y1  y2  0, y 2  y1

Therefore

g ( y1 , y2 , y3 ) ( y1  y2 )( y 2  y3 ) y3e  y1 y1  0,
0  y2  y1 ,
0  y3  y2
Generating function technique

This technique is usually used to determine the probability distribution of


functions of independent random variables.

For this technique, find the generating function for the function of the
random variables using their j.p.d.f and then compare it to known
generating functions to see if there is a match.
Example 5
Let X 1 , X 2 ,..., X n be independent random variables , each with p.d.f

f ( xi ) e  xi xi  0
n

Using the m.g.f technique, find the p.d.f of Z  X i


i 1
n
n
   
M Z (t )   
i 1   t    t 
This is the moment generating function of a gamma distribution with
parameters n and 

Hence the p.d.f of Z is


n
 n  1  z
f ( z)  z e z 0
n
Example 6
Let X 1 , X 2 ,..., X n be independent random variables , each with p.d.f

Pr( X i  xi )  p xi (1  p )1 xi xi 0,1


n
Using the c.f technique, find the p.d.f of Z  X i
i 1
n
Z (t )   X i (t )
i 1

1
 X i (t )   p (1  p)
xi 1 xi
e itxi
(1  p)  pe it

xi 0

n
n
Z (t ) (1  p)  pe  (1  p)  pe 
it it

i 1
n
n
Z (t ) (1  p)  pe   (1  p )  pe 
it it

i 1

This is the characteristic function of a binomial distribution with


parameters n and p

Hence the p.d.f of Z is

 n z n z
Pr( Z  z )   p (1  p ) ; z  0,1, 2, .., n
 z

You might also like