BOOTSTRAP BANDWIDTH SELECTION FOR A SMOOTH SURVIVAL FUNCTION ESTIMATOR FROM CENSORED DATA

Abstract

Since the seminal paper by Efron [5], the bootstrap technique has been applied
extensively to problems in the analysis of censored data. In this paper, the gen-
eral problem is discussed, and a brief literature review is given. A new smooth
estimator for the survival function is proposed, and its strong consistency proved.
A bandwidth selection procedure based on Efron's censored bootstrap that in-
corporates a plug-in principle is proposed to determine a choice for the optimal
bandwidth. Simulations indicates favorable performance of the proposed estima-
tor when compared to the original Kaplan-Meier estimator.