On the assumption of independent right censoring
Seminar in Insurance and Economics
SPEAKER: Morten Overgaard (Aarhus University).
TITLE: On the assumption of independent right censoring.
ABSTRACT: Some sort of assumption of independent right censoring is required for the Kaplan—Meier estimator to be consistent. In this talk, we will take a look at various assumptions made in the literature and see how they can be grouped or categorized. Two categories emerge, one with stronger assumptions than the other. The assumptions of the weaker category are the weakest possible assumptions for the Kaplan—Meier estimator to be consistent on the relevant interval. The assumptions of the stronger category involve allowing for appropriate prognosis after censoring by using the Kaplan—Meier estimator. When also considering a given censoring time, the two categories are divided into six categories depending on whether strong, weak or no further assumptions are made to ensure that the censoring distribution can be consistently estimated by the Kaplan—Meier estimator. All six categories are non-empty as can be demonstrated by simple examples. The practical implications of these categorizations are slight but not nonexistent.