Estimation of the conditional survival function with interval-censored data

Abstract

This paper focuses on the estimation of the conditional survival function in the presence of interval-censored data. While estimation is relatively straightforward for qualitative covariates using Turnbull’s estimator, the case of continuous covariates, particularly under interval censoring, poses additional methodological challenges. We describe existing kernel-based approaches, such as those using Nadaraya-Watson weights, applicable under right censoring. The main contribution of this work is the proposal of two novel methodologies tailored for estimating the conditional survival function when data are interval-censored. These methods use kernel weights and a modified EM algorithm to provide practical ways to solve this challenging problem.