Module 13

Transformations of Absolutely Continuous

Random Variables

() Module 13 Transformations of Absolutely Continuous Random Variables 1 / 15

X : an A.C. r.v. with d.f. FX (·) and p.d.f. fX (·);
Z x
FX (x) = fX (t)dt, x ∈ R;
−∞

For −∞ < a < b < ∞, P ({X = a}) = 0,

P ({a < X ≤ b}) = P ({a < X < b}) = P ({a ≤ X < b})
Z b
= P ({a ≤ X ≤ b}) = fX (t)dt;
a
In general, for any set A ⊆ R,
Z Z ∞
P({X ∈ A}) = fX (t)dt = fX (t)IA (t)dt.
A −∞

() Module 13 Transformations of Absolutely Continuous Random Variables 1 / 15

We will assume throughout that fX (·) is continuous everywhere except
at (possibly) finite number of points (say, x1 , x2 , . . . , xn ), where it has
jump discontinuities. In that case FX (·) is differentiable everywhere
except at discontinuity points {x1 , . . . , xn } of fX (·). Moreover
 0
 FX (t), if t ∈ / {x1 , . . . , xn }
fX (t) = .
0, otherwise


Support

SX = {x ∈ R : FX (x + ) − FX (x − ) > 0, ∀  > 0} .

() Module 13 Transformations of Absolutely Continuous Random Variables 2 / 15

 g : R → R: a given function;

Then Y = g (X ) is a r.v.;

Goal: To find the probability distribution (i.e., d.f. FY (·) and/or,

p.d.f./p.m.f. fY (·)) of Y = g (X );

Remark 1: We have seen that when X is discrete, Y = g (X ) is also

discrete. When X is A.C., Y = g (X ) may not be A.C. (or even
continuous) as the following example illustrates.

() Module 13 Transformations of Absolutely Continuous Random Variables 3 / 15

Example 1: Let X be an A.C. r.v. with p.d.f.
 1
 2 , if − 1 < x < 1
fX (x) = .
0, otherwise


Let Y = [X ] (maximum integer contained in X ). Note that

P({X ∈ (−1, 1)}) = 1.

 −1, if − 1 < X < 0
Y = .
0, if 0 ≤ X < 1


Then
Z 0
P({Y = −1}) = P({−1 < X < 0}) = fX (x)dx
−1
Z 0
1 1
= dx = ,
−1 2 2

() Module 13 Transformations of Absolutely Continuous Random Variables 4 / 15

 Z 1
P({Y = 0}) = P({0 ≤ X < 1}) = fX (x)dx
0
Z 1
1 1
= dx = .
0 2 2

Thus Y is discrete with support SY = {−1, 0} and p.m.f.

 1
 2 , if y ∈ {−1, 0}
fY (y ) = .
0, otherwise


The following result provides sufficient conditions under which a function

of an A.C. random variable is A.C.

() Module 13 Transformations of Absolutely Continuous Random Variables 5 / 15

Result 1: Suppose SX = ki=1 Si,X , where {Si,X , i = 1, . . . , k} is a
S
collection of disjoint intervals and in Si,X (i = 1, ..., k), g : Si,X → R is
strictly monotone with inverse function gi−1 (y ) such that dy d −1
gi (y ) is
continuous. Let g (Si,X ) = {g (x) : x ∈ Si,X }, i = 1, ..., k. Then the r.v.
Y = g (X ) is A.C. with p.d.f.
k  
X
 d −1 
fY (y ) = fX gi−1 (y )  g (y ) I
 dy i  g (Si,X ) (y ),
i=1

where, for a set A, IA (·) denotes its indicator function, i.e.,


1, if y ∈ A
IA (y ) = .
0, otherwise

() Module 13 Transformations of Absolutely Continuous Random Variables 6 / 15

Corollary 1: Suppose that g : SX → R is strictly monotone with inverse
function g −1 (y ) such that dyd −1
g (y ) is continuous. Let
g (SX ) = {g (x) : x ∈ SX }. Then Y = g (X ) is of A.C. type with p.d.f.
 
−1
 d −1 
fY (y ) = fX (g (y ))  g (y ) Ig (SX ) (y ).
dy

() Module 13 Transformations of Absolutely Continuous Random Variables 7 / 15

Example 2: Let X be an A.C. r.v. with p.d.f.
 |x|


 2 , if − 1 < x < 1


fX (x) = x ,
 3, if 1 < x < 2




0, otherwise

and let Y = X 2 .
(a) Find the p.d.f. of Y and hence find the d.f. of Y ;
(b) Find the d.f. of Y and hence find the p.d.f. of Y .

Solution: SX = [−1, 2] = [−1, 0) ∪ [0, 2] = S1,X ∪ S2,X .

g (x) = x 2 , x ∈ SX , is monotone in S1,X and S2,X .

() Module 13 Transformations of Absolutely Continuous Random Variables 8 / 15

 S1,X = [−1, 0) S2,X = [0, 2]
√ √
g1−1 (y ) = − y g2−1 (y ) = y
d −1 −1 d −1 1
dy g1 (y ) = 2 y dy g2 (y ) = 2 y
√ √

g (S1,X ) = (0, 1] g (S2,X ) = [0, 4)

y ∈ g (S1,X ) ⇔ 0 < y ≤ 1 y ∈ g (S2,X ) ⇔ 0 ≤ y ≤ 4
fX (g1−1 (y ))  dy
 d −1 
fX (g2−1 (y ))  dy
 d −1 
g (y ) g2 (y )
 1
√  √ 
 
= fX (− y )  2−1 √  I(0,1] (y ) = fX ( y )  2√1 y  I[0,4] (y )
 
y

Thus a p.d.f. of Y is
2  
X  d −1 
fY (y ) = fX (gi−1 (y ))  g (y ) Ig (Si,X ) (y )
dy i
i=1
1


 2, if 0 < y < 1



1
= 6, if 1 ≤ y < 4 .




0, otherwise

() Module 13 Transformations of Absolutely Continuous Random Variables 9 / 15
 
Z y  0, if y < 0
FY (y ) = P({Y ≤ y }) = fY (t)dt = .
−∞ 
1, if y ≥ 4
For 0 ≤ y < 1,
Z y
1 y
FY (y ) = P({Y ≤ y }) = dt = .
0 2 2
For 1 ≤ y < 4
Z 1 Z y
1 1 y +2
FY (y ) = dt + dt = .
0 2 1 6 6
Thus the d.f. of Y is

 0, if y <0
 y,

if 0≤y <1
FY (y ) = 2 .
y +2
 6 ,
 if 1≤y <4
1, if y ≥4


() Module 13 Transformations of Absolutely Continuous Random Variables 10 / 15

(b) For y < 0,

FY (y ) = P({Y ≤ y }) = P({X 2 ≤ y }) = 0.

For y ≥ 0,

√
y
√ √
Z
2
FY (y ) = P({X ≤ y }) = P({− y ≤ X ≤ y }) = √ X
f (t)dt.
− y

For 0 ≤ y < 1, √
y
|t|
Z
y
FY (y ) = √
dt = .
− y 2 2
For 1 ≤ y < 4,
√
1 y
|t|
Z Z
t
FY (y ) = dt + dt
−1 2 1 3
y +2
= .
6
() Module 13 Transformations of Absolutely Continuous Random Variables 11 / 15
For y ≥ 4, FY (y ) = 1.

Thus the d.f. of Y is


 0, if y <0
 y,

if 0≤y <1
FY (y ) = 2 .
y +2
, if 1≤y <4
 6


1, if y ≥4

() Module 13 Transformations of Absolutely Continuous Random Variables 12 / 15

Clearly SY = [0, 4], FY (·) is differentiable everywhere except at points 0, 1
and 4. Let
 0
FY (y ), if y ∈ / {0, 1, 4}
g (y ) =
0, otherwise
 1

 2 , if 0 < y < 1



1
= 6 , if 1 < y < 4 .




0, otherwise

R∞
Then −∞ g (y )dy = 1. ThusY is of A.C. type with p.d.f.
 1

 2 , if 0 < y < 1



1
fY (y ) = g (y ) = 6 , if 1 < y < 4 .




0, otherwise


() Module 13 Transformations of Absolutely Continuous Random Variables 13 / 15

Take home problem

Let X be a r.v. with p.d.f.

e −x , if x > 0

fX (x) = .
0, otherwise

Let Y = 2X + 3.
(a) Find the p.d.f. of Y and hence find the d.f. of Y;
(b) Find the d.f. of Y and hence find the p.d.f. of Y.

() Module 13 Transformations of Absolutely Continuous Random Variables 14 / 15

Abstract of Next Module

X: a r.v. associated with a random experiment E;

Each time the random experiment is performed we get a value of X ;
Question: If the random experiment is performed infinitely what is the
mean (or expectation) of observed values of X or g (X ), for some function
real-valued function g (·)?
In the discrete case, the relative frequency interpretation of probability
suggests that we take
P
x∈SX g (x) × Number of times we get {X = x}
E (g (X )) = lim
N→∞ N
X frequency of {X = x}
= g (x) lim
N→∞ N
x∈SX
X X
= g (x)P({X = x}) = g (x)fX (x).
x∈SX x∈SX

() Module 13 Transformations of Absolutely Continuous Random Variables 15 / 15