An identifiable model for informative censoring

The usual model for censored survival analysis requires the assumption that censoring of observations arises only due to causes unrelated to the lifetime under consideration. It is easy to envision situations in which this assumption is unwarranted, and in which use of the Kaplan-Meier estimator and associated techniques will lead to unreliable analyses.