Contributions to the study of extreme behaviors and applications

Abstract : The main objective of this thesis is to explore several approaches to deal with extreme behaviors. We start defining a new class M of positive and measurable functions with support R^+ and polynomial asymptotic tail behavior, strictly larger than the class of regularly varying (RV) functions. The functions U∊M are identified by a real index, called the M-index of U, which corresponds to the RV index when U is RV. Algebraic and analytic properties and characterizations of M are given. M is extended into two classes, called M_∞ and M_(-∞), of which functions have exponential asymptotic tail behaviors of the types e^x and e^(-x) respectively. Properties satisfied on M also hold on those classes. Extensions on M of Karamata’s theorem and Karamata’s Tauberian theorem are given. Relations between the domain of attraction of Fréchet and M, as well as that of Gumbel and M_(-∞) are provided. Using a characterization of M, a unified proof of the Tauberian theorems of exponential type given by Kohlbecker, de Bruijn, and Kasahara is given. The second part of the thesis presents on one hand an empirical analysis on the economic benefits generated by the partnership of Swiss Life France with a third-party organization revealing non-linear relations between variables involved in the study; on the other hand, an empirical study on relations between mortality and market risks provides evidence of weak dependence between these extremes. The last part of the thesis presents an accelerated hazard relational model, embedded in a Poisson regression framework, showing an excellent fit to real data.
Document type :
Theses
Complete list of metadatas

Cited literature [130 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01361307
Contributor : Abes Star <>
Submitted on : Wednesday, September 7, 2016 - 10:12:08 AM
Last modification on : Friday, March 22, 2019 - 1:31:03 AM
Long-term archiving on : Thursday, December 8, 2016 - 1:08:15 PM

File

2016PA066038.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01361307, version 1

Citation

Meitner Cadena. Contributions to the study of extreme behaviors and applications. Statistics [math.ST]. Université Pierre et Marie Curie - Paris VI, 2016. English. ⟨NNT : 2016PA066038⟩. ⟨tel-01361307⟩

Share

Metrics

Record views

308

Files downloads

188