Lectures 7 & 8

Transformation Method

Theorem (1): Let 𝑋 be a random variable with 𝑝𝑑𝑓 𝑓𝑋 (𝑥) and let 𝑌 = 𝑢(𝑋) be a

function of 𝑋. Assume that 𝑦 = 𝑢(𝑥) defines a one-to-one

transformation from 𝐴 = {𝑥: 𝑓𝑋 (𝑥) > 0} onto 𝐵 = {𝑦: 𝑔𝑌 (𝑦) >

0} with inverse transformation 𝑥 = 𝑤(𝑦):

1. If 𝑋 is a discrete random variable, then the 𝑝𝑑𝑓 of 𝑌 is,

𝑓 (𝑤(𝑦)) , 𝑦 ∈ 𝐵
𝑔𝑌 (𝑦) = { 𝑋
0 , 𝑜. 𝑤

2. If 𝑋 is a continuous random variable, then the 𝑝𝑑𝑓 of 𝑌 is,

𝑑
𝑓𝑋 (𝑤(𝑦)) | 𝑤(𝑦)| , 𝑦 ∈ 𝐵
𝑔𝑌 (𝑦) = { 𝑑𝑦
0 , 𝑜. 𝑤

𝑑
Provided that 𝑤(𝑦) ≠ 0.
𝑑𝑦

Example 1: Let 𝑋 be a random variable has a geometric 𝑝𝑑𝑓 given by

(1 − 𝑝)𝑥−1 𝑝 , 𝑥 = 1, 2, , 3, …
𝑓𝑋 (𝑥) = {
0 , 𝑜. 𝑤

Use the transformation method to find the 𝑝𝑑𝑓 of 𝑌 = 𝑋 − 1.

Solution:

𝐴 = {𝑥: 𝑓𝑋 (𝑥) > 0} = {𝑥: 𝑥 = 1, 2, , 3, … }

The transformation 𝑦 = 𝑥 − 1 maps 𝐴 onto

𝐵 = {𝑦: 𝑔𝑌 (𝑦) > 0} = {𝑦: 𝑦 = 0, 1, 2, , … }

1
Since 𝑦 = 𝑥⏟− 1 then the inverse transformation is
𝑢(𝑥)
𝑥 = 𝑦⏟+ 1
𝑤(𝑦)
The 𝑝𝑑𝑓 of 𝑌 is,

𝑓 (𝑤(𝑦)) = 𝑓𝑋 (𝑦 + 1), 𝑦 ∈ 𝐵
𝑔𝑌 (𝑦) = { 𝑋
0 , 𝑜. 𝑤

(1 − 𝑝)𝑦+1−1 𝑝, 𝑦∈𝐵
={
0 , 𝑜. 𝑤
(1 − 𝑝)𝑦 𝑝, 𝑦 = 0, 1, 2, , …
={
0 , 𝑜. 𝑤
Example 2: Let 𝑋 be a random variable with 𝑝𝑑𝑓 𝑏𝑖𝑛(𝑛, 𝑝). Find the 𝑝𝑑𝑓 of

𝑌 = 𝑛 − 𝑋.

Solution:

𝑛
( ) 𝑝 𝑥 (1 − 𝑝)𝑛−𝑥 , 𝑥 = 0, 1, 2, … , 𝑛
𝑓𝑋 (𝑥) = { 𝑥
0 , 𝑜. 𝑤

Therefore,

𝐴 = {𝑥: 𝑓𝑋 (𝑥) > 0} = {𝑥: 𝑥 = 0, 1, 2, … , 𝑛}

The transformation 𝑦 = 𝑛 − 𝑥 maps 𝐴 onto

𝐵 = {𝑦: 𝑔𝑌 (𝑦) > 0} = {𝑦: 𝑦 = 0, 1, 2, … , 𝑛}

Since 𝑦 = 𝑛⏟− 𝑥 then the inverse transformation is

𝑢(𝑥)
𝑥 = 𝑛⏟− 𝑦
𝑤(𝑦)
The 𝑝𝑑𝑓 of 𝑌 is,

𝑓 (𝑤(𝑦)) = 𝑓𝑋 (𝑛 − 𝑦), 𝑦 ∈ 𝐵
𝑔𝑌 (𝑦) = { 𝑋
0 , 𝑜. 𝑤

2
 𝑛
(𝑛 − 𝑦) 𝑝𝑛−𝑦 (1 − 𝑝)𝑛−(𝑛−𝑦) , 𝑦 ∈ 𝐵
={
0 , 𝑜. 𝑤
𝑛
(𝑦) 𝑝𝑛−𝑦 (1 − 𝑝)𝑦 , 𝑦 = 0, 1, 2, … , 𝑛
={
0 , 𝑜. 𝑤
That is, 𝑌~𝑏𝑖𝑛(𝑛, 1 − 𝑝).
𝑛 𝑛! 𝑛! 𝑛
Note that: (𝑛 − 𝑦) = = = (𝑦).
(𝑛−(𝑛−𝑦))!(𝑛−𝑦)! 𝑦!(𝑛−𝑦)!

Example 3: Let 𝑋 be a random variable with 𝑝𝑑𝑓

(𝑥) 2𝑒 −2𝑥 , 𝑥>0

𝑓𝑋 ={
0 , 𝑜. 𝑤
Find the 𝑝𝑑𝑓 of 𝑌 = 𝑒 𝑋 .

Solution:
𝐴 = {𝑥: 𝑓𝑋 (𝑥) > 0} = {𝑥: 𝑥 > 0}
The transformation 𝑦 = 𝑒 𝑥 maps 𝐴 onto
𝐵 = {𝑦: 𝑔𝑌 (𝑦) > 0} = {𝑦: 𝑦 > 1}
𝑦 = 𝑒 𝑥 then the inverse transformation is
Since 𝑦 = ⏟
𝑢(𝑥)
𝑥 = 𝑙𝑛
⏟(𝑦)
𝑤(𝑦)
𝑑 𝑑𝑥 1
and 𝑤(𝑦) = = . Therefore, the 𝑝𝑑𝑓 of 𝑌 is,
𝑑𝑦 𝑑𝑦 𝑦
𝑑 1
𝑓𝑋 (𝑤(𝑦)) | 𝑤(𝑦)| = 𝑓𝑋 (𝑙𝑛(𝑦)) | | , 𝑦∈𝐵
𝑔𝑌 (𝑦) = { 𝑑𝑦 𝑦
0 , 𝑜. 𝑤
1
2𝑒 −2𝑙𝑛(𝑦) , 𝑦 ∈ 𝐵
={ 𝑦
0 , 𝑜. 𝑤
2
, 𝑦>1
= {𝑦 3
0 , 𝑜. 𝑤

3
Theorem (2): Let 𝑋1 and 𝑋2 be two random variables with joint 𝑝𝑑𝑓 𝑓(𝑥1 , 𝑥2 )

and let 𝑌1 = 𝑢1 (𝑋1 , 𝑋2 ) and 𝑌2 = 𝑢2 (𝑋1 , 𝑋2 ) are two functions of 𝑋1 and 𝑋2 .

Assume that 𝑦1 = 𝑢1 (𝑥1 , 𝑥2 ) and 𝑦2 = 𝑢2 (𝑥1 , 𝑥2 ) defines one-to-one

transformation from 𝐴 = {(𝑥1 , 𝑥2 ) ∶ 𝑓(𝑥1 , 𝑥2 ) > 0} onto 𝐵=

{(𝑦1 , 𝑦2 ): 𝑔(𝑦1 , 𝑦2 ) > 0} with inverse transformation 𝑥1 = 𝑤1 (𝑦1 , 𝑦2 ) and

𝑥2 = 𝑤2 (𝑦1 , 𝑦2 ):

1. If 𝑋1 and 𝑋2 are discrete random variables, then the joint 𝑝𝑑𝑓 of 𝑌1 and 𝑌2 is,

𝑓[𝑤1 (𝑦1 , 𝑦2 ), 𝑤2 (𝑦1 , 𝑦2 ) ] , (𝑦1 , 𝑦2 ) ∈ 𝐵

𝑔(𝑦1 , 𝑦2 ) = {
0 , 𝑜. 𝑤

2. If 𝑋1 and 𝑋2 are continuous random variables, then the joint 𝑝𝑑𝑓 of 𝑌1 and 𝑌2

is,

𝑓[𝑤1 (𝑦1 , 𝑦2 ), 𝑤2 (𝑦1 , 𝑦2 ) ]|𝐽| , (𝑦1 , 𝑦2 ) ∈ 𝐵

𝑔(𝑦1 , 𝑦2 ) = {
0 , 𝑜. 𝑤

Provided 𝐽 ≠ 0. Where 𝐽 is called the Jacobian and is given by

𝑑𝑤1 (𝑦1 , 𝑦2 ) 𝑑𝑤1 (𝑦1 , 𝑦2 ) 𝑑𝑥1 𝑑𝑥1

𝑑𝑦1 𝑑𝑦2 || = ||𝑑𝑦1 𝑑𝑦2
𝐽 = || |
𝑑𝑤2 (𝑦1 , 𝑦2 ) 𝑑𝑤2 (𝑦1 , 𝑦2 ) 𝑑𝑥2 𝑑𝑥2 |
𝑑𝑦1 𝑑𝑦2 𝑑𝑦1 𝑑𝑦2

𝑎 𝑏
Note that: | | = 𝑎𝑑 − 𝑏𝑐 , which is the determinant of the matrix
𝑐 𝑑

𝑎 𝑏
[ ].
𝑐 𝑑

4
Example 4: The joint 𝑝𝑑𝑓 of 𝑋1 and 𝑋2 is,

2 𝑥1+𝑥2 1 2−𝑥1−𝑥2
𝑓(𝑥1 , 𝑥2 ) = {(3) ( ) , (𝑥1 , 𝑥2 ) = (0,0), (0,1), (1,0), (1,1)
3
0 , 𝑜. 𝑤
1. Find the marginal 𝑝𝑑𝑓 of 𝑋1 .
2. Find the marginal 𝑝𝑑𝑓 of 𝑋2 .
3. Find joint 𝑝𝑑𝑓 of 𝑌1 = 𝑋1 − 𝑋2 and 𝑌2 = 𝑋1 + 𝑋2 .
4. Find the marginal 𝑝𝑑𝑓 of 𝑌1 .
5. Find the marginal 𝑝𝑑𝑓 of 𝑌2 .
Solution: First, note that the range (𝑥1 , 𝑥2 ) = (0,0), (0,1), (1,0), (1,1)

equivalent to 𝑥1 = 0, 1, 𝑥2 = 0, 1.

1. The marginal 𝑝𝑑𝑓 of 𝑋1 is,

𝑓𝑋1 (𝑥1 ) = ∑ 𝑓(𝑥1 , 𝑥2 ) , 𝑥1 = 0, 1

∀𝑥2
𝟏
2 𝑥1+𝑥2 1 2−𝑥1−𝑥2
= ∑( ) ( )
3 3
𝑥2 =𝟎

2 𝑥1 1 2−𝑥1 2 𝑥1+1 1 2−𝑥1−1

=( ) ( ) +( ) ( )
3 3 3 3
2 𝑥1 1 1−𝑥1 1 2
=( ) ( ) ( + )
3 3 3 3
2 𝑥1 1 1−𝑥1
= {(3) (3) , 𝑥1 = 0, 1
0 , 𝑜. 𝑤
2
Note that: 𝑋1 ~𝑏𝑖𝑛(1, )
3

1. The marginal 𝑝𝑑𝑓 of 𝑋2 is,

𝑓𝑋2 (𝑥2 ) = ∑ 𝑓(𝑥1 , 𝑥2 ) , 𝑥2 = 0, 1

∀𝑥1

5
 𝟏
2 𝑥1+𝑥2 1 2−𝑥1−𝑥2
= ∑( ) ( )
3 3
𝑥1 =𝟎

2 0+𝑥2 1 2−0−𝑥2 2 1+𝑥2 1 2−1−𝑥2

=( ) ( ) +( ) ( )
3 3 3 3
2 𝑥2 1 1−𝑥2 1 2
=( ) ( ) ( + )
3 3 3 3
2 𝑥2 1 1−𝑥2
= {(3) (3) , 𝑥2 = 0, 1
0 , 𝑜. 𝑤
2
Note that: 𝑋2 ~𝑏𝑖𝑛(1, )
3

3.

𝐴 = {(𝑥1 , 𝑥2 ): 𝑓(𝑥1 , 𝑥2 ) > 0} = {(𝑥1 , 𝑥2 ): (𝑥1 , 𝑥2 ) = (0,0), (0,1), (1,0), (1,1)}

= {(𝑥1 , 𝑥2 ): 𝑥1 = 0, 1, 𝑥2 = 0, 1}

The transformation 𝑦1 = 𝑥1 − 𝑥2 and 𝑦2 = 𝑥1 + 𝑥2 maps 𝐴 onto

𝐵 = {(𝑦1 , 𝑦2 ): 𝑔(𝑦1 , 𝑦2 ) > 0}

= {(𝑦1 , 𝑦2 ): (𝑦1 , 𝑦2 ) = (0,0), (−1,1), (1,1), (0,2)} (has 4 pairs)

≠ {(𝑦1 , 𝑦2 ): 𝑦1 = −1,0,1, 𝑦2 = 0,1,2} (has 9 pairs)

𝑦1 = 𝑥1 − 𝑥2 and 𝑦2 = 𝑥1 + 𝑥2

𝑦1 = 𝑢1 (𝑥1 , 𝑥2 ) and 𝑦2 = 𝑢2 (𝑥1 , 𝑥2 )

Since 𝑦1 = 𝑥
⏟1 − 𝑥2 and 𝑦2 = 𝑥 ⏟1 + 𝑥2 then the inverse transformation can be
𝑢1 (𝑥1 , 𝑥2 ) 𝑢2 (𝑥1 , 𝑥2 )

obtained as follows,

6
𝑦1 = 𝑥1 − 𝑥2 ⟹ 𝑥2 = 𝑥1 − 𝑦1
𝑦2 = 𝑥1 + 𝑥2 ⟹ 𝑥2 = 𝑦2 − 𝑥1 } ⟹ 𝑥1 − 𝑦1 = 𝑦2 − 𝑥1 ⟹ 𝑥1 = (𝑦
⏟ 1 + 𝑦2 )/2
𝑤1 (𝑦1 , 𝑦2 )

and
𝑥2 = (𝑦
⏟ 2 − 𝑦1 )/2
𝑤2 (𝑦1 , 𝑦2 )

The joint 𝑝𝑑𝑓 of 𝑌1 and 𝑌2 is,

𝑓[𝑤1 (𝑦1 , 𝑦2 ), 𝑤2 (𝑦1 , 𝑦2 ) ] , (𝑦1 , 𝑦2 ) ∈ 𝐵

𝑔(𝑦1 , 𝑦2 ) = {
0 , 𝑜. 𝑤
𝑦1 + 𝑦2 𝑦2 − 𝑦1
𝑓( , ) , (𝑦1 , 𝑦2 ) ∈ 𝐵
={ 2 2
0 , 𝑜. 𝑤
𝑦1+𝑦2 𝑦2 −𝑦1 𝑦1 +𝑦2 𝑦2 −𝑦1
+
2 2 2 1 2− 2
−
2
= {(3) ( ) , (𝑦1 , 𝑦2 ) ∈ 𝐵
3
0 , 𝑜. 𝑤
2 𝑦2 1 2−𝑦2
= { (3) (3) , (𝑦1 , 𝑦2 ) = (0,0), (−1,1), (1,1), (0,2)
0 , 𝑜. 𝑤
4. The marginal 𝑝𝑑𝑓 of 𝑌1 is,

𝑔𝑌1 (𝑦1 ) = ∑ 𝑔(𝑦1 , 𝑦2 ) , 𝑦1 = −1, 0, 1

∀𝑦2

Since the values of 𝑦1 should be −1, 0, 1 then we cannot sum over the

values of 𝑦2 = 0, 1, 2 for each value of 𝑦1 (because in this case we are

adding 9 pairs, not 4). To be correct, if 𝑦1 = −1 then 𝑦2 takes only one

value, which is 𝑦2 = 1. If 𝑦1 = 0 then 𝑦2 = 0, 2 and if 𝑦1 = 1 then 𝑦2 = 1.

Therefore,

7
 𝟏
2 1 1 2−1 2
𝑔𝑌1 (−1) = ∑ 𝑔(−1, 𝑦2 ) = 𝑔(−1,1) = ( ) ( ) =
3 3 9
𝑦2 =1

1 4 5
𝑔𝑌1 (0) = ∑ 𝑔(0, 𝑦2 ) = 𝑔(0,0) + 𝑔(0,2) = + =
9 9 9
𝑦2 =𝟎,𝟐

𝟏
2
𝑔𝑌1 (1) = ∑ 𝑔(1, 𝑦2 ) = 𝑔(1,1) =
9
𝑦2 =1

Thus, the marginal 𝑝𝑑𝑓 of 𝑌1 is summarized in the following table,

𝒚𝟏 -1 0 1 o.w

𝒈𝒀𝟏 (𝒚𝟏 ) 2/9 5/9 2/9 0

5. You can solve this part by the same technique of the previous part.

Example 5: Let 𝑋1 , 𝑋2 be a random sample from 𝑒𝑥𝑝(1) (See also Example 6 of

Lectures 5 and 6).

1. Find the joint 𝑝𝑑𝑓 of 𝑌1 = 𝑋1 + 𝑋2 and 𝑌2 = 𝑋1 − 𝑋2 .

2. Find the marginal 𝑝𝑑𝑓 of 𝑌1 = 𝑋1 + 𝑋2 .

3. Find the marginal 𝑝𝑑𝑓 of 𝑌2 = 𝑋1 − 𝑋2 .

Solution: The joint 𝑝𝑑𝑓 of 𝑋1 and 𝑋2 is,

𝑒 −(𝑥1+𝑥2) , 𝑥1 > 0, 𝑥2 > 0
𝑓(𝑥1 , 𝑥2 ) = {
0 , 𝑜. 𝑤
𝐴 = {(𝑥1 , 𝑥2 ): 𝑓(𝑥1 , 𝑥2 ) > 0} = {(𝑥1 , 𝑥2 ): 𝑥1 > 0, 𝑥2 > 0}

Since 𝑦1 = 𝑥1 + 𝑥2 (= 𝑢1 (𝑥1 , 𝑥2 )) and 𝑦2 = 𝑥1 − 𝑥2 (= 𝑢2 (𝑥1 , 𝑥2 )), then the

inverse transformation is,

8
 𝑦1 −𝑦2 𝑦1 +𝑦2
𝑥1 = (= 𝑤1 (𝑦1 , 𝑦2 )) and 𝑥2 = (= 𝑤2 (𝑦1 , 𝑦2 )) (See Example 4,
2 2

Part3).
To determine the set 𝑩:

We know that 𝑥1 > 0 , 𝑥2 > 0, which gives

𝑦1 = 𝑥1 + 𝑥2 > 0 and −∞ < 𝑦2 = 𝑥1 − 𝑥2 < ∞. In addition, we will try to

discover if the two random variables depends on each other or not,

𝑦1 − 𝑦2
𝑥1 > 0 ⟹ > 0 ⟹ 𝑦1 > 𝑦2
2
𝑦1 + 𝑦2
𝑥2 > 0 ⟹ > 0 ⟹ 𝑦1 > −𝑦2
2
Therefore,

𝐵 = {(𝑦1 , 𝑦2 ): 𝑔(𝑦1 , 𝑦2 ) > 0}

= {(𝑦1 , 𝑦2 ): 𝑦1 > 0, −∞ < 𝑦2 < ∞, 𝑦1 > 𝑦2 , 𝑦1 > −𝑦2 }.
The Jacobian 𝐽 is,
𝑑𝑥1 𝑑𝑥1 1 1
−
𝑑𝑦
𝐽 = || 1
𝑑𝑦2
|| = |2 2| = (1) (1) − (− 1) (1) = 1
𝑑𝑥2 𝑑𝑥2 1 1 2 2 2 2 2
𝑑𝑦1 𝑑𝑦2 2 2
The joint 𝑝𝑑𝑓 of 𝑌1 and 𝑌2 is,

𝑓[𝑤1 (𝑦1 , 𝑦2 ), 𝑤2 (𝑦1 , 𝑦2 )]|𝐽| , (𝑦1 , 𝑦2 ) ∈ 𝐵

𝑔(𝑦1 , 𝑦2 ) = {
0 , 𝑜. 𝑤
𝑦1 − 𝑦2 𝑦1 + 𝑦2 1
= {𝑓 ( , ) , (𝑦1 , 𝑦2 ) ∈ 𝐵
2 2 2
0 , 𝑜. 𝑤
−(
𝑦1 −𝑦2 𝑦1 +𝑦2
+ ) 1
(𝑦1 , 𝑦2 ) ∈ 𝐵
= {𝑒 ,
2 2
2
0 , 𝑜. 𝑤
1 −𝑦
= {2 𝑒
1 , 𝑦1 > 0, −∞ < 𝑦2 < ∞, 𝑦1 > 𝑦2 , 𝑦1 > −𝑦2
0 , 𝑜. 𝑤

9
The set 𝐵 can be represent graphically as given below:

𝒚𝟐

y1=y2

𝒚𝟏

y1=-y2

2. The marginal 𝑝𝑑𝑓 of 𝑌1 is,

𝑔𝑌1 (𝑦1 ) = ∫ 𝑔(𝑦1 , 𝑦2 )𝑑𝑦2 , 𝑦1 > 0

𝑅𝑦2

𝑦1
1 −𝑦
=∫ 𝑒 1 𝑑𝑦2
−𝑦1 2

𝑦 𝑒 −𝑦1 , 𝑦1 > 0
={ 1
0 , 𝑜. 𝑤

10
 3. The marginal 𝑝𝑑𝑓 of 𝑌2 is,

𝑔𝑌2 (𝑦2 ) = ∫ 𝑔(𝑦1 , 𝑦2 )𝑑𝑦1 , −∞ < 𝑦2 < ∞

𝑅𝑦1

∞
1 −𝑦
∫ 𝑒 1 𝑑𝑦1 , 𝑦2 < 0
−𝑦2 2
= ∞
1 −𝑦
∫ 𝑒 1 𝑑𝑦1 , 𝑦2 > 0
{ 𝑦2 2

𝑦
𝑒 2
, 𝑦2 < 0
= { −𝑦22
𝑒
, 𝑦2 > 0
2
|𝑦 |
𝑒 2
= , −∞ < 𝑦2 < ∞
2
---------------------------------------------------------------------------------------------------------

Assignment # 2

Q1: Let 𝑋 be a random variable with 𝑝𝑑𝑓 𝑁(0,1). Find the 𝑝𝑑𝑓 of 𝑌 = 𝑋 2 .

Q2: Let 𝑋1 , 𝑋2 be a random sample from 𝑈(0,1).

1. Use the distribution function method to find the 𝑝𝑑𝑓 of 𝑌1 = 𝑋1 + 𝑋2 .

2. Use the transformation method to find the 𝑝𝑑𝑓 of 𝑌1 = 𝑋1 + 𝑋2 .

Q3: The joint 𝑝𝑑𝑓 of 𝑋1 , 𝑋2 is

0.5 , 0 < 𝑥1 < 1, 0 < 𝑥2 < 2

𝑓(𝑥) = {
0 , 𝑜. 𝑤

Use the transformation method to find the joint 𝑝𝑑𝑓 of 𝑌1 = 𝑋1 + 𝑋2 and

𝑌2 = 𝑋2 .

11