ST 1010: Statistical Theory (2C,30L) DST @ UOC

ST1010: Statistical Theory (2C, 30L)

Function of Random Variables

Let E be an experiment and S be a sample space associated with E. Let X be a random variable defined on S.
Suppose Y = H ( X ) is a real value function of X, then Y = H ( X ) is also a random variable.

Y is a Discrete Functions
Case I : X is discrete and then Y is also discrete
Case II : X is continuous and then Y is discrete

Case I: X is discrete and then Y is also discrete

Suppose that possible values of X are ( x1 , x2 , x3 ,..., xn ) . Than possible values of Y = H ( X ) are;

Y= y1 y2 y3 . . . yn
H(x1) H(x2) H(x3) . . . H(xn)
Here we want to find the probability distribution function of Y as follows:
Y y1 y2 y3 . . . yn
P[Y=y]

Example 01: Suppose that a random variable X has 3 values such that -1, 0, +1 with probabilities 1/3, 1/2,
and 1/6 respectively. Then find the p.d.f. of Y = H ( X ) = 3 X + 1 .
Answer: Here we have

X -1 0 1
P[X=x] 1/3 1/2 1/6
Therefore the possible values of Y are: And the p.d.f. of Y is:
X -1 0 1 Y -2 1 4
Y=3X+1 -2 1 4 P[Y=y] 1/3 1/2 1/6

Example 02: Suppose that a random variable X has following p.d.f.

X -1 0 1
P[X=x] 1/3 1/2 1/6

Let Y = X 2 . Find the p.d.f. of Y.

1 1 1
Answer: P (Y = y ) = P Y = −1 or Y = 1 = P Y = −1 + P Y = 1 = + =
3 6 2
Y 1 0
P[X=x] 1/2 1/2

General procedure: When X is Discrete

If ( x1 , x2 , x3 ,..., xn ) are the possible values of X with P[ X = xi ] = P( xi ) where i=1,2,3,…n and H is a function
such that each value of Y corresponds to one and only one X (and vies versa), which is called 1-1 function,
then the p.d.f. of Y is as follows.
Y= H(x1) H(x2) H(x3) . . . H(xn)
P[Y=y] P(x1) P(x2) P(x3) . . . P(xn)

See example 01

However, in some situations H(X) may not 1-1 function. (That is, y may correspond to more that one x). If
ji
xi1 , xi 2 , xi 3 ,..., xiji all leads to same yi where i = 1, 2,3,..., n then P[Y = yi ] =  P[ X = xik ]
k =1
See example 02

RABA 1
ST 1010: Statistical Theory (2C,30L) DST @ UOC
Case II: X is continuous and then Y is discrete
Example 03: Suppose X is assuming all real values. If we define Y:
1 for x  0
Y =
−1 for x  0

Then P Y = 1 = P[ X  0] =  f ( x)dx
0
0
P Y = −1 = P[ X  0] =  f ( x)dx
−

Y is a Continuous Function
General procedure: When X is Continuous
Step (01): Obtain the cumulative distribution function of Y (That is GY ( y ) )
dGY ( y )
Step (02): Obtain the probability density function of Y (That is gY ( y ) = )
dy
Step (03): Determine the values of y, in the range space for which gY ( y )  0 .

Example 04: Suppose X has p.d.f. f ( x) = 2 x 0  x  1 . Let Y = 3X + 1. Find the p.d.f. of Y.

Answer: Here Y = 3X + 1 is 1-1 function.
y −1 y −1
y −1
y −1 3 3
x2 ( y − 1)2
GY ( y) = P[Y  y] = P[3 X + 1  y] = P[ X 
3
]= 
0
f ( x)dx = 0
2 xdx = 2
2 0
3
=
9
Therefore GY ( y ) is:
 y −1
0 0
3 0 y 1

 ( y − 1) 2 y −1 → 
 ( y − 1) 2
GY ( y ) =  0 1 GY ( y ) =  1 y  4
 9 3  9
 y −1 1 y4
1 1
 3

Therefore the p.d.f. of y is;

dGY ( y ) d  ( y − 1)  2 ( y − 1)
2

gY ( y ) = =  = 1 y  4
dy dy  9  9
 
Note:
4
1) Show that g
1
Y ( y )dy = 1

2) Draw the C.D.F of Y

Example 05: Suppose that the continuous random variable has p.d.f.. f ( x) = 2 x 0  x  1 . Let Y = e − x
. (That is 1-1 function) Find the p.d.f. of Y?

Answer:
GY ( y ) = P[Y  y ] = P[e− x  y ] = P[− x  ln y ] = P[ x  − ln y ]
1 1
x2 1
= 
− ln y
f ( x)dx = 
− ln y
2 xdx =2
2 − ln y

= 1 − ( ln y )
2

RABA 2
ST 1010: Statistical Theory (2C,30L) DST @ UOC
0 − ln y  0

GY ( y ) = 1 − ( ln y )
2
0  − ln y  1
1 − ln y  1


Therefore the p.d.f. of y is gY ( y ) =

dGY ( y) d 1 − (log y)
=
2

=
(
−2log y ) e−1  y  1
dy dy y
Note:
1
1. Show that g −1
Y ( y )dy = 1
e
2. Draw the C.D.F of Y

Theorem : Let X be a continuous random variable with a p.d.f. f ( x ) for a  x  b . Suppose that Y = H ( X )
is a strictly monotone [strictly increasing or decreasing] (and also 1-1) function of X. Assume that H(X) is
differentiable at all x. Then the random variable Y = H ( X ) has a p.d.f. gY ( y ) as follows:
dx
gY ( y) = f X ( x) 
dy
Note: There are two situations for H.
1. Assume that H is a strictly increasing function. Then gY ( y )  0 for values of y such that
H ( a )  y  H (b )

2. Assume that H is a strictly decrease function. Then gY ( y )  0 for values of y such that
H (b )  y  H ( a )

Proof: Consider the above 2 situations separately.

Situation 01: When H is increasing
GY ( y) = P[Y  y] = P[ H ( X )  y] = P[ X  H −1 ( y )] = F ( x ) where x = H −1 ( y )
dGY ( y ) dF ( x ) dx dx
gY ( y ) =
= = f ( x) → (01)
dy dy dy dy
Situation 02: When H is decreasing
GY ( y ) = P[Y  y ] = P[ H ( X )  y ] = P[ X  H −1 ( y )]
= 1 − P[ X  H −1 ( y )] = 1 − F ( x ) where x = H −1 ( y )
dGY ( y ) d 1 − F ( x )  dx dx  dx 
gY ( y ) = = = − f ( x ) = f ( x ) −  → (02)
dy dy dy dy  dy 
From (01) and (02)
dx
gY ( y) = f X ( x)
dy

RABA 3
ST 1010: Statistical Theory (2C,30L) DST @ UOC
Example 06: Suppose X has p.d.f.
f ( x) = 2 x 0  x 1
Let Y = 3X + 1 (1-1 and strictly increasing function).
Find the p.d.f. of Y.

Answer: We know that

dx
gY ( y) = f X ( x)
dy
Here
dx 1
= and f ( x) = 2 x
dy 3
Therefore
1  y − 1  1 2 ( y − 1)
gY ( y ) = 2 x = 2  = for 1  y  4
3  3 3 9

Example 07: Suppose X is a continuous random variable with p.d.f.

1
f ( x) = −1  x  1
2
Let Y = X 2 (not a 1-1 function) . Find the p.d.f. of Y.

Answer:
GY ( y ) = P Y  y  = P  X 2  y  = P  − y  X  + y 
+ y + y

=  f ( x ) dx = 
1
2
1 +
dx =  x −
2
y
y
=
1
2
(
y− − y 
 )
− y − y

= y
Therefore
dGY ( y ) 1
gY ( y ) = = for 0  y  1
dy 2 y

Note: In general
P  a  X  b = F ( b ) − F ( a )
GY ( y) = P − y  X  + y  = FX ( y ) − F (− y ) X

dG ( y ) d 
gY ( y ) = Y
dy
=
dy 
FX ( y ) − F ( − y ) = f ( y ) 2 1 y + f ( − y ) 2 1 y
X X X

=
1 
2 y
fX ( y ) + f ( − y ) = 2 1 y  12 + 12 
X

1
= for 0  y  1
2 y

RABA 4