Bivariate survival function estimation with censored data

Abstract

A family of simple nonparametric estimators is proposed for the underlying bivariate distribution of paired failure times in the presence of bivariate right censoring. The estimators are path-dependent estimators which are generalization of an estimator proposed by Campbell & Foldes (1982). Under univariate censoring, one particular path will produce the estimator with the smallest variance among the family. Under univariate censoring, an algorithm is proposed to obtain a "best" path-dependent estimator.