Estimation of high-dimensional extreme conditional expectiles

Gilles Stupfler 1 Stéphane Girard 2
2 MISTIS - Modelling and Inference of Complex and Structured Stochastic Systems
Inria Grenoble - Rhône-Alpes, LJK - Laboratoire Jean Kuntzmann, INPG - Institut National Polytechnique de Grenoble
Abstract : The concept of expectiles is a least squares analogue of quantiles. It has, over the last decade, received a fair amount of attention due to its potential for application in financial, actuarial, and economic contexts. Some very recent work has focused on the application of extreme expectiles to assess tail risk, and on their estimation in a heavy-tailed framework. We investigate the estimation of extreme conditional expectiles of a heavy-tailed random variable $Y$ given a finite-dimensional covariate $X$, whose dimension is allowed to be large. We derive generic conditions under which the limiting behaviour of our estimators can be investigated. We then present applications of our results to certain regression models of particular interest, as well as a finite-sample study to get a grasp of the behaviour of our procedures in practice.
Type de document :
Communication dans un congrès
11th International Conference of the ERCIM WG on Computing and Statistics, Dec 2018, Pisa, Italy. 2018
Liste complète des métadonnées

https://hal.archives-ouvertes.fr/hal-01942210
Contributeur : Stephane Girard <>
Soumis le : lundi 3 décembre 2018 - 08:52:25
Dernière modification le : vendredi 7 décembre 2018 - 14:41:25

Identifiants

  • HAL Id : hal-01942210, version 1

Collections

Citation

Gilles Stupfler, Stéphane Girard. Estimation of high-dimensional extreme conditional expectiles. 11th International Conference of the ERCIM WG on Computing and Statistics, Dec 2018, Pisa, Italy. 2018. 〈hal-01942210〉

Partager

Métriques

Consultations de la notice

23