28.05.2013 Views

Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...

Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...

Etude des marchés d'assurance non-vie à l'aide d'équilibres de ...

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.

tel-00703797, version 2 - 7 Jun 2012<br />

4.6. Appendix<br />

A large part of this paper focuses on the discrete time framework. We have seen that<br />

discrete time ruin probability asymptotics can be obtained in a similar way as in the continuous<br />

case. However, <strong>de</strong>riving asymptotics for the claim distribution is far more difficult: complex<br />

analysis is necessary to tackle the binomial alternating sum issue. Furthermore, the <strong>non</strong>uniqueness<br />

of discrete copulas is also studied. We quantify the maximal difference between<br />

the continuous and the discrete settings. Despite the issues encountered with discrete copula,<br />

the latent variable approach is consi<strong>de</strong>red in many other articles, e.g., Joe (1997); Frees and<br />

Wang (2006); Channouf and L’Ecuyer (2009); Braeken et al. (2007); Leon and Wu (2010).<br />

As mentioned in Albrecher et al. (2011), the approach proposed in this paper to <strong>de</strong>rive<br />

new explicit formula can be used for more general risk mo<strong>de</strong>ls. It could be interesting to<br />

test whether the A + B/u still applies for the ruin probability, say, with phase-type claim<br />

distributions. Beyond the study of ruin probability, the mixing approach and the asymptotic<br />

rule might probably be used for finite-time ruin probabilities and the Gerber-Shiu function.<br />

4.6 Appendix<br />

4.6.1 Usual special functions<br />

List of common special functions<br />

Let us recall the <strong>de</strong>finition of the so-called special functions, see Olver et al. (2010) for a<br />

comprehensive and updated list. In most cases, the <strong>de</strong>finition does not limit to z ∈ R, but<br />

can be exten<strong>de</strong>d to z ∈ C.<br />

– the gamma function Γ(α) = ∞<br />

0 xα−1 e −x dx,<br />

– the lower incomplete gamma function γ(α, z) = z<br />

0 xα−1 e −x dx,<br />

– the upper incomplete gamma function Γ(α, z) = ∞<br />

z xα−1 e −x dx,<br />

– the beta function β(a, b) = 1<br />

0 xa−1 (1 − x) b−1 dx,<br />

– the incomplete beta function β(a, b, z) = z<br />

0 xa−1 (1 − x) b−1 dx,<br />

– the complementary incomplete beta function ¯ β(a, b, z) = 1<br />

z xa−1 (1−x) b−1 dx = β(b, a, 1−<br />

z),<br />

– the error function erf(z) = 2<br />

z<br />

√<br />

π 0 e−t2dt,<br />

– the complementary error function erfc(z) = 2<br />

– the binomial coefficient n<br />

k =<br />

n!<br />

k!(n−k)! .<br />

Known asymptotics of usual special functions<br />

√ π<br />

+∞<br />

z e−t2dt, In this subsection, we list asymptotics of the main special functions, see Olver et al. (2010).<br />

Gamma function Known as the Stirling formula, the asymptotic expansion for Γ(z) is<br />

Γ(z) ∼ +∞ e −z z z<br />

2π<br />

z<br />

<br />

1 + 1<br />

12z<br />

gk<br />

+ · · · + + . . .<br />

zk for z ∈ C and | arg(z)| < π with g0 = 1, g1 = 1/12, g2 = 1/288, g3 = −139/51480. See page<br />

141 of Olver et al. (2010).<br />

We recall that the incomplete gamma functions are <strong>de</strong>fined as<br />

γ(a, z) =<br />

<br />

,<br />

z<br />

x<br />

0<br />

α−1 e −x ∞<br />

dx, and Γ(a, z) = x<br />

z<br />

α−1 e −x dx.<br />

205

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

Saved successfully!

Ooh no, something went wrong!