0% found this document useful (0 votes)
6 views20 pages

Bivariate Variables and Distributions

This lecture covers bivariate variables, focusing on jointly distributed random variables, including discrete and continuous types. Key concepts such as expected value, covariance, correlation, and conditional distributions are discussed, along with examples like contingency tables. The lecture concludes with an introduction to the bivariate normal distribution and joint moment generating functions.

Uploaded by

Anh Nguyen
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
6 views20 pages

Bivariate Variables and Distributions

This lecture covers bivariate variables, focusing on jointly distributed random variables, including discrete and continuous types. Key concepts such as expected value, covariance, correlation, and conditional distributions are discussed, along with examples like contingency tables. The lecture concludes with an introduction to the bivariate normal distribution and joint moment generating functions.

Uploaded by

Anh Nguyen
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

LECTURE 8

Bivariate Variables

Lecturer: Nguyen Thi Thu Van


Email: [Link]@[Link]
Contents

 Jointly distributed random variables

 Expected value, Covariance, and Correlation

 Conditional distributions

 Bivariate normal distribution

 Joint Moment Generating Function


Jointly distributed random variables
Jointly distributed Random Variables
 There are many experimental situations in
which more than one random variable will be
of interest to an investigator.

 We shall first consider joint probability


distributions for

• two discrete random variables,

• then for two continuous variables,

• and finally for more than two variables.


Two Discrete Random Variables
Let 𝑋 and 𝑌 be two discrete variables defined on the sample space 𝑆 of
an experiment. The joint probability mass function 𝑝(𝑥, 𝑦) is defined for
each pair of numbers (𝑥, 𝑦) by 𝑝 𝑥, 𝑦 = 𝑃 𝑋 = 𝑥 𝑎𝑛𝑑 𝑌 = 𝑦

Let 𝐴 be any set consisting of pairs of (𝑥, 𝑦) values. Then the probability
that the random pair (𝑋, 𝑌) lies in 𝐴 is obtained by summing the joint
pmf over pairs in 𝐴:

𝑃 𝑋, 𝑌 ∈ 𝐴 = ෍ ෍ 𝑝(𝑥, 𝑦)
𝑥,𝑦 ∈𝐴

The marginal probability mass function of 𝑋 and 𝑌, denoted 𝑝𝑋 (𝑥) and


𝑝𝑌 (𝑦), respectively, are given by

𝑝𝑋 𝑥 = σ𝑦 𝑝(𝑥, 𝑦) and 𝑝𝑌 𝑦 = σ𝑥 𝑝(𝑥, 𝑦).


Contingency Table
 Collect data of 100 cars:
 Each car either has AC or no AC

 Each car either has GPS or no GPS

GPS No GPS Total


AC 35 55 90
No AC 5 5 10
Total 40 60 100
Of the 100 cars studied,
 Marginal probability: 90% have air conditioning
(AC) and 40% have a GPS.
 Joint probability:

35% of the cars have both.

GPS No GPS Total


AC 0.35 0.55 0.90
No AC 0.05 0.05 0.10
Total 0.40 0.60 1.00

 Conditional probability:
P(GPS  AC) 0.35
P(GPS | AC)    0.3889
P(AC) 0.90
Two Continuous Random Variables
Let 𝑋 and 𝑌 be two continuous variables. Then f(𝑥, 𝑦) is
a joint probability density function for 𝑋 and 𝑌 if for any
two dimensional set 𝐴

𝑃 𝑋, 𝑌 ∈ 𝐴 = ඵ 𝑓 𝑥, 𝑦 𝑑𝑥𝑑𝑦
𝐴

The marginal probability mass function of 𝑋 and 𝑌,


denoted 𝑓𝑋 (𝑥) and 𝑓𝑌 (𝑦), respectively, are given by

𝑓𝑋 𝑥 = ‫׬‬−∞ 𝑓 𝑥, 𝑦 𝑑𝑦 for −∞ < 𝑥 < ∞

𝑓𝑌 𝑦 = ‫׬‬−∞ 𝑓 𝑥, 𝑦 𝑑𝑥 for − ∞ < 𝑥 < ∞.
Two Independent Continuous Random Variables

Two variables 𝑋 and 𝑌 are said to be independent if


for every pair of 𝑥 and 𝑦 values, we have that

 𝑝 𝑥, 𝑦 = 𝑝𝑋 𝑥 ⋅ 𝑝𝑌 (𝑦) when 𝑋, 𝑌 are discrete

or

 𝑓 𝑥, 𝑦 = 𝑓𝑋 𝑥 ⋅ 𝑓𝑌 (𝑦) when 𝑋, 𝑌 are continuous.

Otherwise, 𝑋 and 𝑌 are said to be called dependent.


Expected value, Covariance, and Correlation
Let 𝑋 and 𝑌 be jointly distributed random variables with pmf 𝑝(𝑥, 𝑦) or pdf
𝑓(𝑥, 𝑦) according to whether the variables are discrete or continuous. Then
the expected value of a function ℎ(𝑋, 𝑌), denoted by 𝐸 ℎ 𝑋, 𝑌 or 𝜇ℎ(𝑋,𝑌) is
given by

෍ ෍ ℎ 𝑥, 𝑦 ⋅ 𝑝 𝑥, 𝑦 𝑋 𝑎𝑛𝑑 𝑌 𝑎𝑟𝑒 𝑑𝑖𝑠𝑐𝑟𝑒𝑡𝑒,


𝑥 𝑦
𝐸 ℎ 𝑋, 𝑌 = ∞ ∞

න න ℎ 𝑥, 𝑦 ⋅ 𝑓 𝑥, 𝑦 𝑑𝑥𝑑𝑦 𝑋 𝑎𝑛𝑑 𝑌 𝑎𝑟𝑒 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠.


−∞ −∞

If two variables are independent,

ℎ 𝑋, 𝑌 = 𝑋 ⋅ 𝑌, 𝐸 𝑋𝑌 = 𝐸 𝑋 × 𝐸(𝑌).

If 𝑋1 , 𝑋2 , . . . , 𝑋𝑛 are independent random variables and assume that the


expected values of ℎ(𝑋1 ), ℎ(𝑋2 ), . . . , ℎ(𝑋𝑛 ) all exist. Then
𝐸 ℎ(𝑋1 ) ℎ 𝑋2 ⋯ ℎ(𝑋𝑛 ) = 𝐸 ℎ(𝑋1 ) 𝐸 ℎ 𝑋2 ⋯ 𝐸 ℎ(𝑋𝑛 )
The covariance of two variables 𝑋 and 𝑌 is

𝐶𝑜𝑣 𝑋, 𝑌 = 𝐸 (𝑋 − 𝜇𝑋 )(𝑌 − 𝜇𝑌 )
෍ ෍ 𝑥 − 𝜇𝑋 𝑦 − 𝜇𝑌 𝑝 𝑥, 𝑦 𝑋 𝑎𝑛𝑑 𝑌𝑑𝑖𝑠𝑐𝑟𝑒𝑡𝑒,
𝑥 𝑦
= ∞ ∞

න න 𝑥 − 𝜇𝑋 𝑦 − 𝜇𝑌 𝑓 𝑥, 𝑦 𝑑𝑥𝑑𝑦 𝑋 𝑎𝑛𝑑 𝑌 𝑐𝑜𝑛𝑡𝑖𝑛𝑢𝑜𝑢𝑠.


−∞ −∞

A shortcut formula for 𝐶𝑜𝑣 𝑋, 𝑌 = 𝐸 𝑋𝑌 − 𝜇𝑋 ⋅ 𝜇𝑌

Meanwhile, we have that


𝐶𝑜𝑣 𝑎𝑋 + 𝑏𝑌, 𝑍 = 𝑎𝐶𝑜𝑣 𝑋, 𝑌 + 𝑏𝐶𝑜𝑣(𝑌, 𝑍)
The correlation of two variables 𝑋 and 𝑌 denoted
𝐶𝑜𝑟𝑟 𝑋, 𝑌 or 𝜌𝑋,𝑌 , or just 𝜌 is defined by

𝐶𝑜𝑣(𝑋,𝑌)
𝜌= .
𝜎𝑋 ⋅𝜎𝑌

By definition, we have that

 If 𝑎, 𝑐 are either both positive or both negative,


𝐶𝑜𝑟𝑟 𝑎𝑋 + 𝑏, 𝑐𝑌 + 𝑑 = 𝐶𝑜𝑟𝑟(𝑋, 𝑌)

 For any two variables 𝑋 and 𝑌, −1 ≤ 𝐶𝑜𝑟𝑟 𝑋, 𝑌 ≤ 1


Conditional distributions
 Let 𝑋 and 𝑌 be two discrete random variables with joint pmf 𝑝(𝑥, 𝑦)
and marginal 𝑋 pmf 𝑝𝑋 (𝑥). Then for any 𝑥 value such that 𝑝𝑋 (𝑥) >
0, the conditional probability mass function of 𝑌 given 𝑋 = 𝑥 is
𝑝(𝑥, 𝑦)
𝑝𝑌|𝑋 𝑦 𝑥 =
𝑝𝑋 (𝑥)
 Similarly, let 𝑋 and 𝑌 be two continuous random variables with joint
pdf 𝑓(𝑥, 𝑦) and marginal 𝑋 pdf 𝑓𝑋 (𝑥). Then for any 𝑥 value such that
𝑓𝑋 (𝑥) > 0, the conditional probability density function of 𝑌 given
𝑋 = 𝑥 is
𝑓(𝑥, 𝑦)
𝑓𝑌|𝑋 𝑦𝑥 =
𝑓𝑋 (𝑥)
 Let 𝑋 and 𝑌 be two discrete random variables with conditional probability
mass function 𝑝𝑌|𝑋 (𝑦, 𝑥). Then the conditional mean or expected value of 𝑌
given that 𝑋 = 𝑥 is

𝜇𝑌|𝑋=𝑥 = 𝐸 𝑦 𝑋 = 𝑥 = ෍ 𝑦𝑝𝑌|𝑋 (𝑦|𝑥)


𝑦∈𝐷𝑌

 Similarly, let 𝑋 and 𝑌 be two continuous random variables with conditional


probability density function 𝑓𝑌|𝑋 (𝑦, 𝑥). Then the conditional mean or
expected value of 𝑌 given that 𝑋 = 𝑥 is

𝜇𝑌|𝑋=𝑥 = 𝐸 𝑦 𝑋 = 𝑥 = න 𝑦𝑓𝑌|𝑋 𝑦 𝑥 𝑑𝑦
−∞
The conditional variance of 𝑌 given that 𝑋 = 𝑥 is
2
𝜎𝑌|𝑋=𝑥 = 𝑉 𝑌 𝑋 = 𝑥 = 𝐸 𝑌 − 𝐸(𝑌|𝑋 = 𝑥) 2 |𝑋 = 𝑥
The conditional mean of any function 𝑔 𝑌 can be obtained similarly.

𝐸 𝑔(𝑦) 𝑋 = 𝑥 = ෍ 𝑔 𝑦 ⋅ 𝑝𝑌|𝑋 (𝑦|𝑥) ; 𝐸 𝑔(𝑦) 𝑋 = 𝑥 = න 𝑔 𝑦 ⋅ 𝑓𝑌|𝑋 𝑦 𝑥 𝑑𝑦


𝑦∈𝐷𝑌 −∞
 If 𝑋 and 𝑌 be two independent discrete random
variables, then
𝑝(𝑥, 𝑦) 𝑝𝑋 𝑥 ⋅ 𝑝𝑌 (𝑦)
𝑝𝑌|𝑋 𝑦 𝑥 = = = 𝑝𝑌 (𝑦)
𝑝𝑋 (𝑥) 𝑝𝑋 (𝑥)
 Similarly, let 𝑋 and 𝑌 be two continuous random
variables, then

𝑓(𝑥, 𝑦) 𝑓𝑋 𝑥 ⋅ 𝑓𝑌 (𝑦)
𝑓𝑌|𝑋 𝑦𝑥 = = = 𝑓𝑌 (𝑦)
𝑓𝑋 (𝑥) 𝑓𝑋 (𝑥)
Bivariate Normal Distribution
Joint Moment Generating Function
-- The End of Topic --
Thank You!

You might also like