25.06.2013 Views

statistique, théorie et gestion de portefeuille - Docs at ISFA

statistique, théorie et gestion de portefeuille - Docs at ISFA

statistique, théorie et gestion de portefeuille - Docs at ISFA

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.

Since the mapping Φ −1 (Fi(·)) is obviously increasing, theorem 2 allows us to conclu<strong>de</strong> th<strong>at</strong> the copula<br />

of the variables yi’s is i<strong>de</strong>ntical to the copula of the variables xi’s. Therefore, the variables yi’s have<br />

Normal marginal distributions and a Gaussian copula with correl<strong>at</strong>ion m<strong>at</strong>rix ρ. Thus, by <strong>de</strong>finition, the<br />

multivari<strong>at</strong>e distribution of the yi’s is the multivari<strong>at</strong>e Gaussian distribution with correl<strong>at</strong>ion m<strong>at</strong>rix ρ :<br />

201<br />

G(y) = Φρ,N(Φ −1 (F1(x1)), · · · , Φ −1 (FN(xN))) (24)<br />

= Φρ,N(y1, · · · , yN), (25)<br />

and y is a Gaussian random vector. From equ<strong>at</strong>ions (24-25), we obviously have<br />

Consi<strong>de</strong>r now the random variable<br />

ρij = Cov[Φ −1 (Fi(xi)), Φ −1 (Fj(xj))]. (26)<br />

z 2 = y t ρ −1 y =<br />

N<br />

yi (ρ −1 )ij yj , (27)<br />

i,j=1<br />

where · t <strong>de</strong>notes the transpose oper<strong>at</strong>or. This variable has already been consi<strong>de</strong>red in (Sorn<strong>et</strong>te <strong>et</strong> al.<br />

2000a) in preliminary st<strong>at</strong>istical tests of the transform<strong>at</strong>ion (23). It is well-known th<strong>at</strong> the variable z 2<br />

follows a χ 2 -distribution with N <strong>de</strong>grees of freedom. In<strong>de</strong>ed, since y is a Gaussian random vector with<br />

covariance m<strong>at</strong>rix 1 ρ, it follows th<strong>at</strong> the components of the vector<br />

˜y = ρ −1/2 y, (28)<br />

are in<strong>de</strong>pen<strong>de</strong>nt Normal random variables. Here, ρ −1/2 <strong>de</strong>notes the square root of the m<strong>at</strong>rix ρ −1 , which<br />

can be obtain by the Cholevsky <strong>de</strong>composition, for instance. Thus, the sum ˜y t ˜y = z 2 is the sum of the<br />

squares of N in<strong>de</strong>pen<strong>de</strong>nt Normal random variables, which follows a χ 2 -distribution with N <strong>de</strong>grees of<br />

freedom.<br />

3.2 Testing procedure<br />

The testing procedure used in the sequel is now <strong>de</strong>scribed. We consi<strong>de</strong>r two financial series (N = 2) of<br />

size T : {x1(1), · · · , x1(t), · · · , x1(T )} and {x2(1), · · · , x2(t), · · · , x2(T )}. We assume th<strong>at</strong> the vectors<br />

x(t) = (x1(t), x2(t)), t ∈ {1, · · · , T } are in<strong>de</strong>pen<strong>de</strong>nt and i<strong>de</strong>nticaly distributed with distribution F ,<br />

which implies th<strong>at</strong> the variables x1(t) (respectively x2(t)), t ∈ {1, · · · , T }, are also in<strong>de</strong>pen<strong>de</strong>nt and<br />

i<strong>de</strong>ntically distributed, with distributions F1 (respectively F2).<br />

The cumul<strong>at</strong>ive distribution ˆ Fi of each variable xi, which is estim<strong>at</strong>ed empirically, is given by<br />

ˆFi(xi) = 1<br />

T<br />

T<br />

1 {xi

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

Saved successfully!

Ooh no, something went wrong!