STS3402

CHAPTER 7
DISTRIBUTION FOR FUNCTIONS OF
RANDOM VARIABLES
TRANSFORMATION TECHNIQUE FOR
DISCRETE RANDOM VARIABLE
◦ Let 𝑋 be a discrete random variable with pdf 𝑝(𝑥). Let 𝑌 = 𝑢(𝑋) define a
one-to-one transformation between the values of 𝑋 and 𝑌, so that the equation
𝑦 = 𝑢(𝑥) can be uniquely solved for 𝑥 in terms of 𝑦 , where 𝑥 =
𝑢−1 𝑦 . Then, 𝑌 is also a random variable and the pdf of 𝑌 is:

𝑝 𝑦 =𝑃 𝑌=𝑦
=𝑃 𝑢 𝑋 =𝑦
= 𝑃(𝑋 = 𝑢−1 𝑦 )
EXAMPLE
X is a random variable with 𝑝 𝑥 = 𝑝𝑞 𝑥−1 , 𝑥 = 1,2,3, …
Determine the distribution of 𝑌 = 𝑋 2 .
Solution:
Discrete case in 2 random variables

◦ Let 𝑋1 and 𝑋2 be a discrete random variables with joint probability distribution

function 𝑝(𝑥1 , 𝑥2 ). If 𝑌1 = 𝑢1 𝑋1 , 𝑋2 and 𝑌2 = 𝑢2 𝑋1 , 𝑋2 where 𝑢1 and 𝑢2
are one-to-one transformations between the points (𝑥1 , 𝑥2 ) and 𝑦1 , 𝑦2 .
Then,

◦ 𝑥1 = 𝑢1−1 𝑦1 , 𝑦2
◦ 𝑥2 = 𝑢2−1 𝑦1 , 𝑦2 (obtained by solving 𝑦1 and 𝑦2 simultaneously)

𝑝 𝑦1 , 𝑦2 = 𝑃 𝑌1 = 𝑦1 , 𝑌2 = 𝑦2
= 𝑝[𝑋1 = 𝑢1−1 𝑦1 , 𝑦2 , 𝑋2 = 𝑢2−1 𝑦1 , 𝑦2 ]
EXAMPLE
𝑋1 and 𝑋2 are independent poisson variables with parameters 𝜆1 and 𝜆2 . Find
𝑝 𝑦1 , 𝑦2 if 𝑌1 = 𝑋1 + 𝑋2 and 𝑌2 = 𝑋2 . Hence, obtain the marginal distribution of 𝑌1 =
𝑋1 + 𝑋2 .
Solution:
Solution:
Solution:
TRANSFORMATION TECHNIQUE FOR
CONTINUOUS RANDOM VARIABLE
 THEOREM

◦ Let 𝑋 be a continuous random variable having pdf 𝑓𝑥 (𝑥). Suppose that 𝑌 = 𝑔(𝑋) be a
function of 𝑋 that is a strictly monotone (increasing or decreasing). Then 𝑌 = 𝑔(𝑋) is
also a random variable and has probability density function,

𝛿𝑔−1 𝑦
𝑓𝑦 𝑦 = 𝑓𝑥 [𝑔−1 𝑦 ]
𝛿𝑦

where 𝑔−1 𝑦 is defined to equal that value of 𝑥 such that 𝑔 𝑥 = 𝑦.

Proof when g(x) is an
increasing function

Note:

If g(x) is an increasing function, then when

𝑔 𝑥 ≤ 𝑦, 𝑥 ≤ 𝑔−1 𝑦
◦𝑌=𝑔 𝑋
◦ 𝐹𝑦 𝑦 = 𝑃 𝑔 𝑋 ≤ 𝑦 = 𝑃 𝑋 ≤ 𝑔−1 𝑦 = 𝐹𝑥 𝑔−1 𝑦

◦ Differentiation gives that

𝛿𝑔−1 𝑦
𝑓𝑦 𝑦 = 𝑓𝑥 𝑔−1 𝑦
𝛿𝑦

◦ Since 𝑔−1 𝑦 is an increasing function, its derivation is non-negative.

Proof when g(x) is a
decreasing function

Note:
If g(x) is a decreasing function, then when
𝑔 𝑥 ≤ 𝑦, 𝑥 ≥ 𝑔−1 𝑦
◦ So, 𝐹𝑦 𝑦 = 𝑃 𝑌 ≤ 𝑦 = 𝑃 𝑔 𝑋 ≤ 𝑦 = 𝑃 𝑋 ≥ 𝑔−1 𝑦 = 1 − 𝑃[𝑋 ≤
𝑔−1 𝑦 ]

◦ Differentiation gives that,

𝛿𝑔−1 𝑦
𝑓𝑦 𝑦 = 𝑓𝑥 𝑔−1 𝑦
𝛿𝑦

𝛿𝑔−1 𝑦
◦ Note: <0, because 𝑔−1 𝑦 is a decreasing function.
𝛿𝑦
EXAMPLE
Let 𝑋 be a continuous r.v with
𝑥2
𝑓 𝑥 = ,1 < 𝑥 < 2
3
If 𝑌 = 𝑋 2 , find 𝑓𝑦 (𝑦).

Solution:
EXAMPLE
𝑓 𝑥 = 2𝑥, 0 < 𝑥 < 1
Find the pdf of 𝑌 = −4𝑥 + 3

Solution:
JOINT PROBABILITY DISTRIBUTIONS OF
FUNCTIONS OF RANDOM VARIABLES
Let X1 and X2 be jointly continuous random variables with joint pdf f(x1, x2). Suppose 𝑌1 =
𝑔1 𝑥1 , 𝑥2 and 𝑌2 = 𝑔2 𝑥1 , 𝑥2 for some functions 𝑔1 , 𝑔2.

Assume 𝑔1 , 𝑔2satisfy the following:

Equations 𝑦1 = 𝑔1 𝑥1 , 𝑥2 and 𝑦2 = 𝑔2 𝑥1 , 𝑥2 can be uniquely solved for x1 and x2 in

terms of y1 and y2 with solutions given by 𝑥1 = ℎ1 𝑥1 , 𝑥2 ; 𝑥2 = ℎ2 𝑥1 , 𝑥2

𝑔1 and 𝑔2 have continuous partial derivations at all points (x1, x2) such that the determinant:
𝜕𝑥1 𝜕𝑥1
𝜕𝑦1 𝜕𝑦2
𝐽 𝑥1 , 𝑥2 = ≠ 0.
𝜕𝑥2 𝜕𝑥2
𝜕𝑦1 𝜕𝑦2
Under these conditions, the random variables Y1 and Y2 are jointly continuous with
joint pdf, 𝑓𝑦 𝑦1 , 𝑦2 = 𝑓𝑥 ℎ1 𝑦1 , 𝑦2 ; ℎ2 𝑦1 , 𝑦2 𝐽 𝑥1 , 𝑥2

𝐽 𝑥1 , 𝑥2 or simplify J is the Jacobian of the transformation.

EXAMPLE
Let X1 and X2 be jointly continuous random variable with pdf,
𝑓 𝑥1 , 𝑥2 = 2𝑥1 ; 0 ≤ 𝑥1 , 𝑥2 ≤ 1. Find the density function of Y = X1 + X2.

Solution:
Solution:
Transformation region
Transformation region
To determine the marginal density, f1(y1), two regions are considered:

1) when 0 ≤ 𝑦1 ≤ 1

24
2) when 1 ≤ 𝑦1 ≤ 2
Hence,

The transformation region 𝑦1 , 𝑦2 becomes

To find the pdf of y1,
◦ when 0 ≤ 𝑦1 ≤ 1

◦ when 1 ≤ 𝑦1 ≤ 2

27
EXAMPLE
X1 and X2 are random variables with joint pdf,
𝑓 𝑥1 , 𝑥2 = 𝑒 − 𝑋1 +𝑋2 ; 0 ≤ 𝑋1 , 𝑋2 ≤ ∞.
Determine the pdf of X1 + X2 .
Solution:
y2
The pdf of X1 + X2 is the marginal density of Y1,

30
The marginal pdf of Y2 is

1)when 0 ≤ 𝑦2 ≤ ∞

2) when −∞ ≤ 𝑦2 ≤ 0
So, the marginal pdf of Y2 is

32
 OTHER METHODS OF
DETERMINING DENSITIES OF
FUNCTIONS OF RANDOM
VARIABLES
 Theorem (Uniqueness)
Let MX(t) and MY(t) be the mgf of X
The Moment and Y if
Generating Function
Method (MGF) MX(t) = MY(t)
for all t, then X and Y are identically
distributed.
EXAMPLE
Xi ; i = 1, 2, …, n are independent exponential random variables,
𝑓 𝑥𝑖 = 𝜆𝑒 −𝜆𝑥𝑖 ; 𝑖 = 1, . . . , 𝑛
Determine the distribution of 𝑍 = σ𝑛𝑖=1 𝑋𝑖 .

Solution:
EXAMPLE
Xi ; i = 1, 2, …, n are independent and normally distributed with 𝐸 𝑋𝑖 = 𝜇𝑖 and
𝑉𝑎𝑟 𝑋𝑖 = 𝜎𝑖2 . Determine the distribution of 𝑍 = σ𝑛𝑖=1 𝑎𝑖 𝑋𝑖 .

Solution:
 1. Xi ~ exponential and are independent random
variables
𝑛

𝑍 = ෍ 𝑋𝑖 ~𝛤 𝜆, 𝑛
Some 𝑖=1

Transformation 2. 𝑋𝑖 ~𝑁 𝜇𝑖 , 𝜎𝑖2 and are independent random

Distribution variables
𝑛

𝑍 = ෍ 𝑎𝑖 𝑋𝑖 ~𝑁 𝜇, 𝜎 2
𝑖=1

𝜇 = σ𝑛𝑖=1 𝑎𝑖 𝜇𝑖 and 𝜎 2 = σ𝑛𝑖=1 𝑎𝑖2 𝜎𝑖2

 3. Zi ~ N(0,1) and are independent random
Some variables then
Transformation 𝑛

Distribution 𝑊 = ෍ 𝑍𝑖2 ~𝜒𝑛2

𝑖=1
 𝑡𝑍 2
PROOF: Zi ~ N(0,1)  𝑀 𝑍2 𝑡 =𝐸 𝑒 =
1
∞ 1 𝑡𝑧 2 − 𝑧2
‫׬‬−∞ 2𝜋 𝑒 ×𝑒 2 𝑑𝑧
1 ∞ −1 1−2𝑡 𝑧 2
= 2𝜋
‫׬‬−∞
𝑒 2 𝑑𝑧

Let 𝑢2 = 1 − 2𝑡 𝑧 2
𝑑𝑢
𝑢= 1 − 2𝑡 𝑧 ⇒ 𝑑𝑧 =
1−2𝑡
1
1 ∞ 1 − 𝑢2
𝑀𝑍 2 𝑡 = 2𝜋
‫׬‬−∞ 1−2𝑡
𝑒 2 𝑑𝑢
1
1 ∞ 1 − 𝑢2
= ‫׬‬ 𝑒 2 𝑑𝑢
1−2𝑡 −∞ 2𝜋

1 1 1/2
= = .
1−2𝑡 1−2𝑡
𝑊 = σ𝑛𝑖=1 𝑍𝑖2
where Zi s are independent
𝑛/2
𝑛 1
𝑀𝑊 𝑡 = 𝛱𝑖=1 𝑀𝑍 2 𝑡 =
1 − 2𝑡
which is the mgf of 𝜒𝑛2 .
4) Zi ~ N(0,1) and 𝑊~𝜒𝜈2 are independent random variables, then

𝑍
𝑇= ~𝑡𝜈
𝑊
𝜈
Proof: Show in assignment/SCL 2
 SUMMARY OF CHAPTER 7
By the end of this chapter students should be familiar with the following:

• Transformations of univariate random variables, discrete and continuous

• Transformations of multivariate random variables, discrete and
continuous
• Methods of determining marginal densities of functions of random
variables