Bias-corrected estimation for conditional Pareto-type distributions with random right censoring (Q2322840)

This paper is concerned with a bias-reduced estimation of the extreme value index for conditional Pareto-type distributions with random right censoring. Let \(Y\) be the response variable of interest, and \(C\) be the random right censoring time, both having a distribution depending on a random covariate \(X\). Assume that conditionally on \(X\) the random variables \(Y\) and \(C\) are independent and the conditional distribution of \(Y\) and \(C\) given \(X=x\) belong to some Pareto-type class \(\mathcal{D}\). Under these assumptions, the authors propose the local likelihood estimation method and prove convergence in probability and asymptotic normality of the obtained estimators. The finite sample behaviour of the estimators is examined by a simulation experiment. The proposed methodology is illustrated by real data examples: AIDS survival data and insurance loss data.