23.01.2014 Views

Modelling Dependence with Copulas - IFOR

Modelling Dependence with Copulas - IFOR

Modelling Dependence with Copulas - IFOR

SHOW MORE
SHOW LESS

Create successful ePaper yourself

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

2.4 <strong>Copulas</strong> and Random Variables<br />

Since X 1 ,X 2 ,... ,X n are continuous, RanG 1 =RanG 2 = ... =RanG n = I. Hence<br />

it follows that C α = C in I n .<br />

From Sklar’s theorem we know that the copula function, C, separates an n-<br />

dimensional distribution function from its univariate margins. The next theorem<br />

will show that there is also a function, Ĉ, that separates an n-dimensional survival<br />

function from its univariate survival margins. Furthermore this function can be<br />

shown to be a copula, and this survival copula can quite easily be expressed in<br />

terms of C and its k-margins.<br />

Theorem 2.7. Let X 1 ,X 2 ,... ,X n be continuous random variables <strong>with</strong> copula<br />

C X1,X 2,... ,X n<br />

.Letα 1 ,α 2 ,... ,α n be strictly monotone on RanX 1 , RanX 2 ,... ,RanX n ,<br />

respectively, and let α 1 (X 1 ),α 2 (X 2 ),... ,α n (X n ) have copula<br />

C α1(X 1),α 2(X 2),... ,α n(X n). Furthermore let α k be strictly decreasing for some k, where<br />

1 ≤ k ≤ n. Without loss of generality let k =1. Then<br />

C α1(X 1),α 2(X 2),... ,α n(X n)(u 1 ,u 2 ,... ,u n )=<br />

= C α2(X 2),... ,α n(X n)(u 2 ,... ,u n ) −<br />

C X1,α 2(X 2),...α n(X n)(1 − u 1 ,u 2 ,... ,u n ).<br />

Proof. Let X 1 ,... ,X n have margins F 1 ,... ,F n and let α 1 (X 1 ),... ,α n (X n )have<br />

margins G 1 ,... ,G n .Then<br />

C α1(X 1),α 2(X 2),... ,α n(X n)(G 1 (x 1 ),... ,G n (x n )) =<br />

= P[α 1 (X 1 ) ≤ x 1 ,... ,α n (X n ) ≤ x n ]<br />

= P[X 1 >α −1<br />

1 (x 1),α 2 (X 2 ) ≤ x 2 ,... ,α n (X n ) ≤ x n ]<br />

= {P[A c ∩ B] =P[B] − P[A ∩ B]}<br />

= P[α 2 (X 2 ) ≤ x 2 ,... ,α n (X n ) ≤ x n ] −<br />

P[X 1 ≤ α −1<br />

1 (x 1),α 2 (X 2 ) ≤ x 2 ,... ,α n (X n ) ≤ x n ]<br />

= C α2(X 2),... ,α n(X n)(G 2 (x 2 ),... ,G n (x n )) −<br />

C X1,α 2(X 2),... ,α n(X n)(F 1 (α −1<br />

1 (x 1),G 2 (x 2 ),... ,G n (x n ))<br />

= {G 1 (x) =P[α 1 (X 1 ) ≤ x] =P[X 1 >α −1<br />

1 (x)] = 1 − F 1(α −1<br />

1 (x))}<br />

= C α2(X 2),... ,α n(X n)(G 2 (x 2 ),... ,G n (x n )) −<br />

C X1,α 2(X 2),... ,α n(X n)(1 − G 1 (x 1 ),G 2 (x 2 ),... ,G n (x n )),<br />

from which the conclusion follows directly.<br />

By using the result of the two theorems above recursively it is clear that the copula<br />

C α1(X 1),α 2(X 2),... ,α n(X n) can be expressed in terms of the copula C X1,X 2,... ,X n<br />

. This<br />

is done in the following example.<br />

Example 2.3. Consider the bivariate case.<br />

Let α 1 be strictly decreasing and let α 2 be strictly increasing. Then<br />

C α1(X 1),α 2(X 2)(u 1 ,u 2 ) = u 2 − C X1,α 2(X 2)(1 − u 1 ,u 2 )<br />

= u 2 − C X1,X 2<br />

(1 − u 1 ,u 2 ).<br />

Let α 1 and α 2 be strictly decreasing. Then<br />

C α1(X 1),α 2(X 2)(u 1 ,u 2 ) = u 2 − C X1,α 2(X 2)(1 − u 1 ,u 2 )<br />

= u 2 − ( 1 − u 1 − C X1,X 2<br />

(1 − u 1 , 1 − u 2 ) )<br />

= u 1 + u 2 − 1+C X1,X 2<br />

(1 − u 1 , 1 − u 2 ).<br />

7

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

Saved successfully!

Ooh no, something went wrong!