Probability Theory and Distribution (Math 2204) Lecture note

Chapter 2. Functions of Random Variables

2.1 Equivalent Events
An Example
Suppose that the radius X of the opening of a finely calibrated tube is considered as a continuous
rv. with pdf f . Let A  X 2 be the cross-sectional area of the opening. It is intuitively clear that
since the value of X is the outcome of a random experiment, the value of A also is. That is, A is a
continuous rv., and we might wish to obtain its pdf, say g .

We would expect that since A is a function of X, the pdf g is in some way derivable from
knowledge of the pdf f .

We shall be concerned with problems of this general nature in this chapter.

Before we familiarize ourselves with some of the specific techniques needed, let us formulate the
above concepts more precisely.

S RX RY

X X(s) =x H
s H(x) =y

Figure 2.1

Equivalent Events
Let £ be an experiment and let S be a sample space associated with £. Let X be a rv. defined on
S . Suppose that y  H (x) is a real –valued function of x . Then Y  H ( X ) is a rv. since for
every s  S , a value of Y is determined, say y  H [ X ( s )] .

Schematically we have Figure2.1.

As before, we call RX the range space of X, the set of all possible values of the function X.
Similarly we define RY as the range space of the rv. Y, the set of all possible values of Y.

Definition: Let C be an event (subset) associated with the range space of Y, RY , as described
above. Let B  RX be defined as follows:

B  x  RX : H ( x)  C . (2.1)

In words: B: is the set of all values of X such that H ( x)  C . If B and C are related in this way
we call them equivalent events.

Meseret Taddesse Ejeta Page 1

Probability Theory and Distribution (Math 2204) Lecture note

Notes
(a) As before, the informal interpretation of the above is that B and Care equivalent events iff B
and C occur together. That is, when B occurs, C occurs and conversely.

(b) Suppose that A is an event associated with S which is equivalent to an event B associated

with RX . Then, if C is an event associated with RY which is equivalent to B, we have that A is

equivalent to C.
(c) It is again important to realize that when we speak of equivalent events (in the above sense),
these events are associated with different sample spaces.

Example 2.1 Suppose that H ( x)   x 2 as in Section 2.1. Then the events B : X  2 and
C : Y  4 are equivalent. For if Y  X 2 , then X  2 occurs iff Y  4  occurs, since X
cannot assume negative values in the present context.(See Fig.2.2)

Fig 2.2
x

Definition: Let X be a rv. defined on the sample space S . Let RX be the range space of X.

Let H be a real-valued function and consider the rv. Y  H ( X ) with range space RY .

For any event C  RY , we define P (C ) as follows:

P(C)  Px  RX : H ( x)  C. (2.2)

In words: The probability of an event associated with the range space of Y is defined as the
probability of the equivalent event (in terms of X) as given by Eq. (2.2).

Meseret Taddesse Ejeta Page 2

Probability Theory and Distribution (Math 2204) Lecture note

Example 2.2: Let X be a continuous rv. with pdf f ( x)  e  x , x  0 .


(A simple integration reveals that  e  x dx  1 ,)
0

Suppose that H ( x)  2 x  1 .

Hence RX  x | x  0 , while RY  y | y  1.

Suppose that the event C is determined as follows:

C  Y  5.

Now y  5 iff 2x  1  5 which in turn yields x  2 .

Hence C is equivalent to B  X  2. (See Fig 2.3)


1
Now P( X  2)   e  x dx  .
2
e2

Hence applying Eq.(2.2) we find that

1
P (Y  5)  .
e2

Figure 2.3

Meseret Taddesse Ejeta Page 3

Probability Theory and Distribution (Math 2204) Lecture note

2.2 Discrete Random Variables

Case 1: X is a discrete rv.
If X is a discrete rv. and Y  H ( X ) ,then it follows immediately that Y is also a discrete rv.

If x1 , x2 ,..., xn ,... are the possible values of X, p( xi )  P( X  xi ) and H is a function such that to
each value y there corresponds exactly one value x, then the probability distribution of Y is
obtained as follows.

Possible values of Y: yi  H ( xi ) , , i  1,2,.., n,...;

Probabilities of Y: q( yi )  P(Y  yi )  p( xi ) .

Example 2.3 Suppose that the random variable X assumes the three values  1,0, and 1 with
1 1 1
probabilities , and , respectively. Let Y  3X  1 . Then the possible values of Y are
3 2 6
1 1 1
 2,1 and 4 ,assumed with probabilities , and .
3 2 6

Case 2:
Quite often the function H does not possess the above characteristic, and it may happen that
several values of X lead to the same value of Y, as the following example illustrates.
Example 2.4: Suppose that we consider the rv. X having three values -1,0 and 1 with
1 1 1
probabilities , and respectively.
3 2 6

1 1
If Y  X 2 , then the possible values of Y are zero and one, assumed with probabilities , .
2 2

1 1 1
For Y  1 iff X  1 or X  1 and the probability of this latter event is   .
3 6 2

In terms of our previous terminology the events B : X  1 and C : Y  1 are equivalent
events and hence Eq.(2.2) have equal probabilities.
The general procedure for situations as described in the above example is as follows:

Let xi1 , xi 2 ..., xik ,..., represent the X-values having the property

H ( xij )  yi for all j . Then

q( yi )  P(Y  yi )  p( xi1 )  p( xi 2 )  ...

Meseret Taddesse Ejeta Page 4

Probability Theory and Distribution (Math 2204) Lecture note

In words: To evaluate the probabilities of the event Y  yi , find the equivalent event in terms of
X (in the range space RX ) and then add all the corresponding probabilities. (See Fig.2.4)

RY


 Fig.2.4


n
1
Example 2.5: Let X have possible values 1,2,..., n,.. and suppose that P ( X  n)  .
2

1 if X is even
Let Y  
 1 if X is odd

Find the probability distribution of Y.

Solution: Possible values of Y are -1 and +1.

Then, since Y  1 iff X  2 , or X  4 , or X  6 , or , ... ,

P(Y  1)  P( X  2)  P( X  4)  ...
1 1 1 1
    ... 
4 16 64 3
2
Hence P (Y  1)  1  P(Y  1)  .
3
Therefore, the probability distribution of Y is given in the table below.

yi 1 1
P(Y  yi ) 2 1
3 3

Meseret Taddesse Ejeta Page 5

Probability Theory and Distribution (Math 2204) Lecture note

2.3 Continuous Random Variables

The most important (and most frequently encountered) arises when X is a continuous rv. with
pdf f and H is a continuous function. Hence Y  H ( X ) is a continuous rv. and it will be our task
to obtain its pdf, say g .

The general procedure will be as follows:

(a) Obtain G, the cdf of Y, where G ( y )  P (Y  y ) , by finding the event A (in the range space of
X) which is equivalent to the event Y  y.

(b) Differentiate G(y) with respect to y in order to obtain g(y).

(c) Determine those values of y in the range space of Y for which g ( y )  0 .

Example 2.6: Suppose that X has pdf

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

Let H ( x)  3x  1 . Find the pdf of Y  H (x) .

Solution y

G ( y )  P (Y  y ) y
 P (3 X  1  y )

 P X  ( y  1) 
 3
( y 1)
3
  2 x dx
0

 ( y  1) 
2
x
 3 

2
Thus, g ( y )  G( y )  ( y  1) . Figure 2.5
9

Since f ( x)  0 for 0  x  1 , we find that g ( y )  0 for 1  y  4 .


Note: The events A, referred to above, equivalent to the event Y  y is simply X  ( y  1) .
3


Meseret Taddesse Ejeta Page 6

Probability Theory and Distribution (Math 2204) Lecture note

OR: Consider again

 y 1
G ( y )  P(Y  y )  P X  
 3 
 y 1
 F 
 3 

Where F is the cdf of X; that is F ( x)  P ( X  x) .

In order to evaluate the derivative of G. G ( y ) , we use the chain rule for differentiation as
follows.

dG( y ) dG( y ) du y 1 g(y)

 . where u  .
dy du dy 3

1
Hence , G( y )  F (u ).
3
1
 q (u ).
3
 y 1 1
 2 . , y
 3 3
y=1 y=4
=
Figure 2.6
as before.
4
The pdf of Y has the graph shown in Fig.2.6 (To check the computation note that  g ( y) dy  1 .)
1

Example 2.7: Suppose that a continuous rv. has pdf as given by

f ( x)  2 x , 0  x  1 .Let H ( x)  e  x .

Find the pdf of Y.

Solution:


G ( y )  P(Y  y )  P e  X  y 
 P( X   ln y )
1
  2 x dx
 ln y

 1  ( ln y ) 2

Meseret Taddesse Ejeta Page 7

Probability Theory and Distribution (Math 2204) Lecture note

 2 ln y
Hence g ( y )  G( y )  .
y
1
Since f ( x)  0 for 0  x  1 , we find that g ( y )  0 for  y 1 .
e

Note that the algebraic sign for g(y) is correct since lny < 0 for 1/e < y < 1. The graph of g(y) is
sketched in Fig.2.8.
y g(y)

x y
x = -lny y = 1/e y=1

Figure 2.7 Figure 2.8

Again we can obtain the above result by slightly different approach which we briefly outline.
As before
G ( y )  P (Y  y )  P ( X   ln y )
 1  P ( X   ln y )  1  F ( ln y ) ,

where F is the cdf of X.

Thus using chain rule we have:

dG( y ) dG du
 . , where u   ln y .
dy du dy

 1
Thus G( y )   F (u )(1 / y )  2 ln y.   , as before.
 y 

Meseret Taddesse Ejeta Page 8

Probability Theory and Distribution (Math 2204) Lecture note

Let us now generalize the approach suggested by the above examples. The crucial step in each of
the examples was taken when we replaced the event Y  y by the equivalent event in terms of
the random variable X. In the above problems this was relatively easy since in each case the
function was a strictly increasing or strictly decreasing function of X.

In Fig.2.9, y is a strictly increasing function of x. Hence we may solve y  H (x) for x in terms of
y, say x  H 1 ( y) ,where H 1 is called the inverse function of H. Thus if H is strictly increasing,
H ( X )  y  
is equivalent to X  H 1 ( y ) while if H is strictly decreasing, H ( X )  y is
 
equivalent to X  H 1 ( y ) .

Figure 2.9

The method used in the above example may now be generalized as follows.

Theorem 2.1: Let X be a continuous random variable with pdf f ,where f ( x)  0 for a  x  b .

Suppose that y  H (x) is a strictly monotone (increasing or decreasing) function of x.

Assume that this function is differentiable (and hence continuous) for all x.

Then the random variable Y defined as Y  H (x) has a pdf g given by,

dx
g ( y )  f ( x) ,
dy

where x is expressed in terms of y. If H is increasing, then g is nonzero for those values of y

satisfying H ( x)  y  H (b). If H is decreasing, then g is nonzero for those values of y satisfying
H (b)  y  H (a ) .

Meseret Taddesse Ejeta Page 9

Probability Theory and Distribution (Math 2204) Lecture note

Proof:
(a) Assume that H is a strictly increasing function. Hence
G ( y )  P(Y  y )  P( H ( x)  y )
 P( X  H 1 ( y ))
 F ( H 1 ( y ))

Differentiating G(y) with respect to y, we obtain, using the chain rule

for derivatives,

dG( y ) dG( y ) dx
 . , where x  H 1 ( y ) .
dy dx dy

dF ( x) dx dx
Thus G( y )  .  f ( x) . .
dx dy dy

(b) Assume that H is a decreasing function.

Therefore
G ( y )  P(Y  y )  P( H ( X )  y )
 P( X  H 1 ( y ))
 1  P( X  H 1 ( y ))
 1  F ( H 1 ( y )

Proceeding as above, we may write

.  1  F ( x)
dG( y ) dG( y ) dx d dx

dy dy dy dx dy
dx
  f ( x). .
dy

Note: The algebraic sign obtained in (b) is correct since, if y is a decreasing function of x, x is a
dx
decreasing function of y and hence  0 . Thus, by using the absolute value sign around dx/dy,
dy
we may combine the result of (a) and (b) and obtain the final form of the theorem.

Meseret Taddesse Ejeta Page 10

Probability Theory and Distribution (Math 2204) Lecture note

Example 2.8: Let us reconsider Example 2.6 and 2.7, by applying Theorem 2.1.

(a) For Example 2.6 we had f ( x)  2 x , 0  x  1 , and y  3 x  1 . Hence x  ( y  1) and

3
dx  . 1
dy 3

Thus g ( y )  2( y  1)  .  ( y  1) , 1  y  4 , which checks the result obtained previously.

1 2
 3  3 9

1
(b) In Example 2.7 we had f ( x)  2 x , 0  x  1 , and y  e  x . .Hence x   ln y and dx  .
dy y
Thus g ( y )   2(ln y ) , 1  y  1 which again checks with the above result.
y e

Example 2.9: Let X be a continuous rv. with pdf

x
 , 1 x  5
f ( x)  12
0 , elsewhere

Find the pdf of the rv. Y  2 X  3 .

Solution: The inverse solution of y  2 x  3 yields x  ( y  3) , from which we obtain

2
dx 1
 . Therefore, using theorem 2.1, we find the density function of Y to be
dy 2
 ( y  3)
 2 1 y3
g ( y )   12  2   48 ,  1  y  7

0 , elsewhere

If y  H (x) is not a monotone function of x we cannot apply the above method directly.

Instead, we shall return to the general method outlined above.

The following example illustrates this procedure.

Example 2.10: Suppose that

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

Let H ( x)  x 2 . This is obviously not a monotone function over the interval [-1, 1]. (Fig.2.10)

Meseret Taddesse Ejeta Page 11

Probability Theory and Distribution (Math 2204) Lecture note
y

x
-1 1
Figure 2.10

g(y)

(1, 1/2)
y
Figure 2.11
Hence we obtain the pdf of Y  X 2 as follows:

G ( y )  P(Y  y )  P( X 2  y )

P yX  y
 F  y   F  y  ,

Where F is the cdf of the rv. X.

Therefore

Meseret Taddesse Ejeta Page 12

Probability Theory and Distribution (Math 2204) Lecture note

g ( y )  G( y ) 
f  y   f  y 
2 y 2 y


2 y
1
 f  y  f  y  .
1 1 1 1
Thus g ( y )     , 0  y  1 . (See Fig2.11)
2 y 2 2 2 y

The method used in the above example yields the following general result.

Theorem 2.2: Let X be a continuous random variable with pdf f . Let Y  X 2 . Then the
random variable Y has pdf given by

g ( y) 
2 y
1
 f  y  f  y  .
Example 2.11: Suppose that X is uniformly distributed over (-1, 1). Let Y  4  X 2 .

Find the pdf of Y, say g(y) and sketch it.

Solution:

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

is pdf of X.

G ( y )  P(Y  y )  P(4  X 2  y )
 P( X 2  4  y )
 1  P( X 2  4  y )

1 P  4  y  X  4  y 
 1  F   
4 y F  4 y 
Therefore,

g ( y )  G( y ) 
f  4 y   f  4 y 
1 1
2(4  y ) 2
2(4  y ) 2


1
1
f   
4 y  f  4 y 
2(4  y ) 2

Meseret Taddesse Ejeta Page 13

Probability Theory and Distribution (Math 2204) Lecture note

1 1 1 1
Thus, g ( y )     , 3 y  4
 2 2  2(4  y ) 2
1 1
2(4  y ) 2

Example 2.12: (a) Given f ( x)  e  x , x  0 , find the probability density function of the variable
Y  X2.

(b) If X has a uniform distribution in [0,1], find the density function of Y  2 ln X .

Solution:

(a) Since y  x 2 is an increasing function of x when x  0 theorem 2.1 can be applied.

dx 1
Hence x  y   .
dy 2 y

1
 g ( y)  e y . , y 0.
2 y

(b) In this case f ( x)  1 , 0  x  1 .

Thus , G ( y )  P(Y  y )  P(2 ln X  y )

 1 
 P ln X   y 
 2 
  y
1
 P X  e 2 
 
 1  P X  e 2 
y

 
y
e 2

 1  1. dx
0
y
 1 e 2

d 1 y
 g ( y)  G( y)  e 2 , y  0 .
dx 2

Meseret Taddesse Ejeta Page 14