23.01.2014 Views

Modelling Dependence with Copulas - IFOR

Modelling Dependence with Copulas - IFOR

Modelling Dependence with Copulas - IFOR

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

3 <strong>Dependence</strong><br />

Since the dependence structure among random variables is represented by copulas<br />

it provides a natural way to study and measure dependence and association<br />

between random variables. Many of these properties and measures are invariant<br />

under strictly increasing transforms (as a direct consequence of Theorem 2.6). Linear<br />

correlation (or Pearson’s correlation) is often used in practice as a measure of<br />

dependence. However since linear correlation is not a copula based dependence<br />

measure, it is often quite misleading and should not be taken as the canonical dependence<br />

measure.<br />

We begin by presenting linear correlation, and then we continue <strong>with</strong> some copula<br />

based measures of dependence.<br />

3.1 Linear Correlation<br />

Definition 8. Let X and Y be two real valued random variables <strong>with</strong> finite variances.<br />

The linear correlation coefficient between X and Y is<br />

ρ l (X, Y )=<br />

Cov(X, Y )<br />

√<br />

Var(X)<br />

√<br />

Var(Y )<br />

,<br />

where Cov(X, Y )=E(XY ) − E(X)E(Y ) is the covariance between X and Y ,and<br />

Var(X),Var(Y ) denotes the variances of X and Y .<br />

Linear correlation is a measure of linear dependence. In the case of perfect<br />

linear dependence, i.e., Y = aX + b almost surely for a ∈ R\{0},b ∈ R, wehave<br />

ρ l (X, Y )=±1. Otherwise, −1

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

Saved successfully!

Ooh no, something went wrong!