A Stratified Product Limit Estimator for the Follow-up Study in Cardio-Thoracic Surgery

We propose a stratified product limit estimator and compare the asymptotic results with those of the unstratified version. When the censoring mechanisms are unequal for different strata, the unstratified version may overestimate the total survival rate of a heterogeneous population. A numerical example in cardiac surgery is examined to demonstrate that this situation does occur in real applications. This overestimation also can be elucidated heuristically by some redistribution schemes for the censored data.