Bias-corrected estimation for conditional Pareto-type distributions with random right censoring

Yuri Goegebeur (1) , Armelle Guillou (2) , Jing Qin (1)

1 IMADA - Department of Mathematics and Computer Science [Odense]
2 IRMA - Institut de Recherche Mathématique Avancée

We consider bias-reduced estimation of the extreme value index in conditional Pareto-type models with random covariates when the response variable is subject to random right censoring. The bias-correction is obtained by fitting the extended Pareto distribution locally to the relative excesses over a high threshold using the maximum likelihood method. Consistency and asymptotic normality of the estimators are established under suitable assumptions. The finite sample behaviour is illustrated with a small simulation experiment and the method is applied to AIDS survival data.

Mots clés: Pareto-type model; random covariate; random right censoring; local estimation; bias-reduction

Domaines: Statistiques [math.ST]

Yuri Goegebeur, Armelle Guillou, Jing Qin. Bias-corrected estimation for conditional Pareto-type distributions with random right censoring. Extremes, 2019, 22, pp.459-498. ⟨10.1007/s10687-019-00341-7⟩. ⟨hal-01826112⟩