01.03.2013 Views

Applied Statistics Using SPSS, STATISTICA, MATLAB and R

Applied Statistics Using SPSS, STATISTICA, MATLAB and R

Applied Statistics Using SPSS, STATISTICA, MATLAB and R

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

A.8.2 Moments<br />

A.8 Multivariate Distributions 425<br />

The moments of multivariate r<strong>and</strong>om variables are a generalisation of the previous<br />

definition for single variables. In particular, for bivariate distributions, we have the<br />

central moments:<br />

k<br />

j<br />

m = Ε[<br />

( X − µ ) ( Y − µ ) ] . A. 44<br />

kj<br />

x<br />

The following central moments are worth noting:<br />

y<br />

m20 = 2 σ X : variance of X; m02 = 2 σ Y : variance of Y;<br />

m = Ε XY − µ µ .<br />

m11 ≡ σXY= σYX: covariance of X <strong>and</strong> Y, with 11 [ ] X Y<br />

For multivariate d-dimensional distributions we have a symmetric positive<br />

definite covariance matrix:<br />

⎡ 2<br />

σ<br />

⎤<br />

1 σ 12 K σ 1d<br />

⎢<br />

2 ⎥<br />

= ⎢σ<br />

21 σ 2 K σ 2d<br />

Σ ⎥ . A. 45<br />

⎢ K K K K ⎥<br />

⎢<br />

⎥<br />

2<br />

⎢⎣<br />

σ d1<br />

σ d 2 K σ d ⎥⎦<br />

The correlation coefficient, which is a measure of linear association between X<br />

<strong>and</strong> Y, is defined by the relation:<br />

σ XY<br />

ρ ≡ ρ XY = . A. 46<br />

σ . σ<br />

X<br />

Y<br />

Properties of the correlation coefficient:<br />

i. –1 ≤ ρ ≤ 1;<br />

ii. ρ XY = ρYX<br />

;<br />

iii. ρ = ±1 iff ( Y − µ Y ) / σ Y = ± ( X − µ X ) / σ X ;<br />

iv. ρ aX + b,cY + d = ρ XY , ac > 0; ρ aX + b,cY + d = −ρ<br />

XY , ac < 0.<br />

If m11 = 0, the r<strong>and</strong>om variables are said to be uncorrelated. Since<br />

Ε [ XY ] = Ε[<br />

X ] Ε[<br />

Y ] if the variables are independent, then they are also<br />

uncorrelated. The converse statement is not generally true. However, it is true in<br />

the case of normal distributions, where uncorrelated variables are also independent.<br />

The definitions of covariance <strong>and</strong> correlation coefficient have a straightforward<br />

generalisation for the d-variate case.<br />

A.8.3 Conditional Densities <strong>and</strong> Independence<br />

Assume that the bivariate r<strong>and</strong>om vector [X, Y] has a density function f(x, y). Then,<br />

the conditional distribution of X given Y is defined, whenever f(y) ≠ 0, as:

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!