# Lecture 3: Combinatorics Those who ignore Statistics are condemned to reinvent Brad Efron Lecture 15: Expectation for Multivariate Distributions Probability Theory and Applications Fall 2008 Outline

Correlation Expectations of Functions of R.V. Covariance Covariance and Independence Algebra of Covariance Correlation Intuition Covariance is a measure of how much RV vary together. Wifes Age and Husbands Age Correlation .97

Example from http://cnx.org/content/m10950/latest/ Sometimes not so perfect Arm Strength Versus Grip Strength Pearsons Correlation R=.63 Negative Correlation Child Labor versus GDP Extreme Correlation 1 Linear relation with positive slope

Extreme: Correlation -1 Linear relation with negative slope Zero Correlation Independent Random Guess Covariance??? Positive, Negative, 0 Crime Rate, Housing Price

SAT Scores, GPA Freshman Year Weight and SAT Score Average Daily Temperature, Housing Price GDP, Infant Mortality Life Expectancy, Infant Mortality Expectations of Functions of R.V. E ( g ( x, y )) g ( x, y ) f ( x, y )dydx

E ( X Y ) ( x y ) f ( x, y )dxdy x f ( x, y )dy dx y f ( x, y )dx dy

xf x ( x)dx yf y ( y )dy E ( X ) E (Y )

NOTE substitute appropriate summation for discrete Variance and Covariance Univariate becomes variance var( X ) E ( X X ) 2 ( x X ) 2 f X ( x) x Multivariate becomes covariance cov( X , Y ) E ( X X )( y Y ) ( x X )( y Y ) f ( x, y ) x, y

Note: var( X ) cov( X , X ) NOTE substitute appropriate integral for continuous Calculating Covariance Can simplify cov( X , Y ) E ( X X )(Y Y ) xy y X xY X Y f ( x, y ) x, y xy f ( x, y ) y X x Y f ( x, y ) X Y x, y

x, y E ( XY ) X Y X Y X Y E ( XY ) E ( X ) E (Y ) Correlation of X and Y Definition cov( X , Y ) ( X ,Y ) XY The correlation always falls in [ -1, 1] It a measure of the linear relation between X

and Y Extreme Cases If X=Y then =1. If X=-Y then =-1. If X and Y independent, then =0. If X=-2Y then =?. Example 2 x 0, y 0, x y 1 f ( x, y ) o.w. 0

Joint is Find correlation of X and Y 1 x 1 y fY ( y ) 2dx 2(1 y ) 0 y 1 f X ( x) 2dy 2(1 x) 0 x 1 0 0

1 1 E (Y ) 2 y (1 y ) dy 1/ 3 E ( X ) 2 x(1 x) dx 1/ 3 0 0 1 1

E (Y ) 2 y (1 y ) dy 1/ 6 E ( X ) 2 x 2 (1 x) dx 1/ 6 var(Y ) 1/ 6 1/ 9 1/18 var( X ) 1/ 6 1/ 9 1/18 2 2 0

2 0 Example Joint is 2 x 0, y 0, x y 1 f ( x, y ) o.w. 0 Find correlation of X and Y

1 1 y 1 E ( X , Y ) 2 xydxdy y (1 y ) 2 dy 1/12 0 0 0 cov( X , Y ) 1/12 1/ 9 1/ 36 cov( X , Y ) 1/ 36 ( X ,Y )

1/ 2 XY 1/18 1/18 Properties of Covariance a) var( X ) cov( X , X ) b) cov(aX+bY)=ab[cov(X,Y)] E (aXbY ) abE ( XY ) E (aX ) aE ( X ) E (bY ) bE (Y ) cov(aX bY ) E (aXbY ) E (aX ) E (bY ) ab[ E ( X , Y ) E ( X ) E (Y )] ab cov( X , Y ) Properties of Covariance

c) var(aX ) a 2 var( X ) d) (aX bY ) sign(a * b) ( X , Y ) ab cov( X , Y ) (aX bY ) ( X ,Y ) | a | X | b | Y ab ( X , Y ) sign(a * b) ( X , Y ) | a || b | Properties of Covariance e) If X and Y are independent cov( X , Y ) 0

correl ( X , Y ) 0 Proof: E ( X , Y ) xyf x ( x) f y ( y )dxdy xf x ( x)dx yfY ( y )dy E ( X ) E (Y ) cov( X , Y ) E ( XY ) E ( X ) E (Y ) 0 Find Covariance X\Y 0 1

-1 0 .3 0 .4 0 1

0 .3 Are X and Y independent X\Y 0 1 -1

0 .3 0 .4 0 1 0

.3 Note Cov(X,Y)=0 does not imply independence of X and Y Independence of X and Y implies cov(X,Y)=0 In this case Y=X2 so the variables are definitely not independent but their covariance is 0 because they have no linear relation. Algebra of variance/covariance/correlation

Given: E ( X ) Y var( X ) X2 E (Y ) Y var(Y ) Y2 cov( X , Y ) XY Calculate mean of Z=2X-3Y+5 variance of Z=2X-3Y+5 Long steps E ( X ) 2 X 3Y 5

Z E ( Z ) 2( X X ) 3(Y Y ) [ Z E ( Z )]2 4( X X ) 2 3(Y Y ) 2 12( X X )(Y Y ) E [ Z E ( Z )]2 4 var( X ) 8 var(Y ) 12cov( X , Y ) Working Rules for linear combinations Z 2 X 3Y 5 Write formula 2 X 3Y Discard Constants 2 2 4 X

9 Y 12 XY Square it Replace squared R.V 4 var( X ) 9 var(Y ) 12cov( X , Y ) with var and crossterms with cov Example Given var(X)=4 var(Y)=10 (X,Y)=1/2 Find variance of X-5Y+6?

Given same facts as previous problem Find covariance x-5y+6 and -4X+3Y+2 Working rule works also for more than two variables Find variance of W=2x-3Y+5Z+1