This preview shows pages 1–2. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: (ii) The correlation coefficient is a number between 1 and +1 this follows again from the CauchySchwarz inequality. (iii) If Y = aX + b holds with a = 0 , then Var( Y ) = a 2 Var( X ) , E ( XY ) = aE ( X 2 ) + bE ( X ) and E ( X ) E ( Y ) = a ( E ( X ) ) 2 + bE ( X ) , thus Cov( X, Y ) = a Var( X ) . It follows that ( X, Y ) = 1 for a > and ( X, Y ) = 1 for a < . Linear dependence between X and Y hence implies that ( X, Y ) attains one of the extreme values. Theorem 4.17. Let ( X, Y ) be a twodimensional random variable with Var( X ) > and Var( Y ) > . (i) If X and Y are independent, then E ( X Y ) = E ( X ) E ( Y ) and ( X, Y ) = 0 . (ii) If ( X, Y ) = 1 or, if ( X, Y ) = 1 , then there exist real numbers a and b such that P ( Y = aX + b ) = 1 . The coefficient a has the same sign as ( X, Y ) . (iii) The mean squared error E ( ( Y aX b ) 2 ) of a random variable Y from a linear function aX + b of the random variable X is minimal if a = Cov( X, Y ) Var( X ) and b = E ( Y ) aE ( X ) . In this case it is given by E ( ( Y aX b ) 2 ) = ( 1 ( X, Y ) ) Var( Y ) . Proof. The proof of (i) will only be given under the addition assumption that either both random variables are discrete or both are continuous. If X and Y are discrete random variables, then E ( XY ) = ij x i y j p ij = ij x i y j p i p j = i x i p i j y j p j = E ( X ) E ( Y ) . If X and Y are continuously distributed with densities f 1 and f 2 , then by f ( x, y ) = f 1 ( x ) f 2 ( y ) , ( x, y ) R 2 , a density f of ( X, Y ) is given. Then: E ( XY ) =   x y f 1 ( x ) f 2 ( y ) d y d x =  x f 1 ( x )d x  y f 2 ( y )d y = E ( X ) E ( Y ) . Let us now show (iii). We find E ( ( Y aX b ) 2 ) = Var( Y aX b ) + ( E ( Y aX b ) ) 2 = Var( Y ) + a 2 Var( X ) 2 a Cov( X, Y ) + ( E ( Y ) aE ( X ) b ) 2 ....
View
Full
Document
This note was uploaded on 05/26/2010 for the course STAT 443 taught by Professor Yuliagel during the Winter '09 term at Waterloo.
 Winter '09
 YuliaGel
 Correlation, Correlation Coefficient

Click to edit the document details