Examples on Transformations of Random Variables

1. Let X U ([, ]). Find the distribution of the random variable Y =

cos X.
The density of X is given by
1

2 if x [, ]
fX (x) =
0 otherwise

Method 1: Note that the range of random variable Y is [1, 1]. There
are two solutions to the equation y = cos x for x [, ], one in [, 0]
and the other in [0, ]. Hence, the density of Y = cos X is given by

X dx
fY (y) = fX (cos y)
1

cos x=y
dy

X 1
= fX (cos y)
1
cos x=y
sin (cos 1 y)
2 1
2 sin (cos1 y) if cos1 y [0, ]
=
0 otherwise
Thus, 1
if y [1, 1]
sin (cos1 y)
fY (y) = (1)
0 otherwise
p
If we further use the fact that sin(cos1 (y)) = 1 y 2 we obtain the
following expression for the PDF:
1
(
if y [1, 1]
fY (y) = 1y 2 (2)
0 otherwise

Method 2: The CDF of X is

xa x+ x 1
FX (x) = = = +
ba 2 2 2
The sets that are equivalent to the event {Y y} are {X < cos1 (y)}
and {X > cos1 (y)}. The CDF of Y is given by
FY (y) = P (Y y) = P (cos X y)

0 y [, 1)
= P (X cos1 y) + P (X cos1 y) if y [1, 1]
1 otherwise


0 y [, 1)
= cos 1
y
1 if y [1, 1]
1 otherwise

1
 assuming that cos1 y 0. The probability density function of Y is
obtained as the derivative of this CDF expression.
2. Square law : Let X U ([1, 1]). Find the distribution of the random
variable Y = X 2 .
The PDF of X is given by
1

2 if x [1, 1]
fX (x) =
0 otherwise
Method 1: Note that the range of random variable Y is [0, 1]. There are
two solutions to the equation y = x2 . Hence, the density of Y = X 2 is
given by

X dx
fY (y) = fX (x)
dy
x2 =y

1 1 1 1
= +
2 2 y 2 2 y
1

2 y if y [0, 1]
= (3)
0 otherwise
Method 2: The CDF of X is
x
FX (x) =
2
The CDF of Y is given by
FY (y) = P (Y y) = P (X 2 y)

0 y [, 0)

= P ( y X y) if y [0, 1]
1 otherwise


0 y [, 1)
y

y
= 2 = y if y [1, 1]
2
1 otherwise

Hence, the density of Y is given by (2).

3. Full-wave rectifier: Let X U ([1, 1]). Find the distribution of the
Y = Xu(X).
The density of X is given by
1

2 if x [1, 1]
fX (x) =
0 otherwise
Note that this transformation is not differentiable so the gradient method
is not applicable in this case. Instead let us look at the set equivalence
method.

2
 For y < 0, the event {Y y} does not have a solution on the real
line and hence reduces to a null event. Consequently the probability
of this event is 0.
For y = 0, the event {Xu(X) 0} has only one solution X = 0 and
the probability of this event is FX (0+ ) FX (0 ).
For y > 0, the event that {Xu(X) y} reduces to the event {X y}
and the probability of this event is just FX (y).
The CDF of the transformed random variable can then be summarized as:

0 y<0
FY (y) = FX (0+ ) FX (0 ) y = 0
FX (y) y>0

Since the input distribution is uniform this reduces to:

0 y0
FY (y) =
FX (y) y > 0