Bandwidth choice for the smooth Kaplan–Meier estimator when the censoring variable can be discontinuous

Abstract

In this paper a new cross-validation method (CVN) for selecting the bandwidth of the smooth Kaplan-Meier estimator is introduced. Its performance is compared to that of the usual cross validation by simulations. The censoring variables in the experimentations are discrete. A set of ideas explaining how the classical and smooth Kaplan-Meier estimators work on the truncation model (usually encountered in reliability studies) is also given in the paper.