Skip to Main content Skip to Navigation
Theses

Modélisation semi-paramétrique des extrêmes conditionnels

Abstract : The main goal of this thesis is to propose new estimators of the tail-index as well as the conditional extreme quantiles in a family of heavy-tailed distributions. The considered family of distributions is defined from a regression model with a location function a(·) and a scale function b(·) which are unknown. The real random variable of interest Y is simultaneously recorded with a deterministic covariate x. The residuals Z of the model are independent of the covariate and their cumulative distribution function belongs to the Fréchet domain of attraction whose the tail-index γ is unknown and assumed to be constant. For more flexibility than purely parametric approaches, we opt for a semi-parametric estimation approach. Also, the constancy of the tail-index allows us to obtain, in the case of small samples, more reliable estimates than in certain purely non-parametric approaches existing in the literature. We establish the asymptotic properties of our estimators and present some results allowing to appreciate their finite sample properties both on simulated and real data.
Document type :
Theses
Complete list of metadata

Cited literature [171 references]  Display  Hide  Download

https://hal.inria.fr/tel-02941714
Contributor : Stephane Girard Connect in order to contact the contributor
Submitted on : Thursday, September 17, 2020 - 5:39:22 PM
Last modification on : Wednesday, March 10, 2021 - 3:04:03 PM
Long-term archiving on: : Thursday, December 3, 2020 - 9:23:56 AM

File

Thèse.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-02941714, version 1

Collections

Citation

Aboubacrène Ag Ahmad. Modélisation semi-paramétrique des extrêmes conditionnels. Statistiques [math.ST]. Université Gaston Berger de Saint-Louis (Sénégal), 2020. Français. ⟨tel-02941714⟩

Share

Metrics

Record views

203

Files downloads

265