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.

9.2. Estim<strong>at</strong>ion du coefficient <strong>de</strong> dépendance <strong>de</strong> queue 325<br />

B Proofs of the corollaries<br />

B.1 First corollary<br />

Corollary 1<br />

If the random variable Y has a rapidly varying distribution function, then λ = 0.<br />

Proof : L<strong>et</strong> us write<br />

For a rapidly varying function ¯ FY , we have<br />

t PY (tx)<br />

FY<br />

¯ (t) = t PY (tx)<br />

¯<br />

FY (tx)<br />

∀x > 1, lim<br />

t→∞<br />

FY (tx)<br />

FY (t)<br />

¯<br />

·<br />

¯<br />

¯FY (tx)<br />

¯FY (t)<br />

. (81)<br />

= 0, (82)<br />

while the leftmost factor of the right-hand-si<strong>de</strong> of equ<strong>at</strong>ion (81) remains boun<strong>de</strong>d as t goes to<br />

infinity, so th<strong>at</strong><br />

t PY (tx) FY<br />

¯ (tx)<br />

lim<br />

t→∞ FY<br />

¯<br />

·<br />

(tx) FY<br />

¯<br />

= f(x) = 0 . (83)<br />

(t)<br />

Since f(x) = 0, we can apply lemma 2 without the hypothesis H3, which conclu<strong>de</strong>s the proof. <br />

B.2 Second corollary<br />

Corollary 2<br />

L<strong>et</strong> Y be regularly varying with in<strong>de</strong>x (−α), and assume th<strong>at</strong> hypothesis H3 is s<strong>at</strong>isfied. Then, the<br />

coefficient of (upper) tail <strong>de</strong>pen<strong>de</strong>nce is<br />

λ =<br />

1<br />

<br />

max<br />

1, l<br />

β<br />

where l <strong>de</strong>notes the limit, when u → 1, of the r<strong>at</strong>io FX −1 (u)/FY −1 (u).<br />

α , (84)<br />

Proof : Karam<strong>at</strong>a’s theorem (see Embrechts, Kluppelberg, and Mikosh (1997, p 567)) ensures th<strong>at</strong><br />

H1 is s<strong>at</strong>isfied with f(x) = α<br />

xα+1 , which is sufficient to prove the corollary. To go one step further,<br />

l<strong>et</strong> us <strong>de</strong>fine<br />

where L1(·) and L2(·) are slowly varying functions.<br />

¯Fy(y) = y −α · L1(y), (85)<br />

¯Fε(ε) = ε −α · L2(ε), (86)<br />

Using the proposition st<strong>at</strong>ed in Feller (1971, p 278), we obtain, for the distribution of the variable<br />

X<br />

¯FX(x) ∼ x −α<br />

<br />

β α for large x.<br />

<br />

x<br />

· L1 + L2(x) ,<br />

β<br />

(87)<br />

Assuming now, for simplicity, th<strong>at</strong> L1 (resp. L2) goes to a constant C1 (resp. C2), this implies th<strong>at</strong><br />

H3 is s<strong>at</strong>istified, since<br />

FX<br />

l = lim<br />

u→1<br />

−1 (u)<br />

FY −1 <br />

= β 1 +<br />

(u) C2<br />

βα 1<br />

α<br />

C1<br />

This allows us to obtain the equ<strong>at</strong>ions (10) and (13). <br />

26<br />

. (88)

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

Saved successfully!

Ooh no, something went wrong!