Censoring, Truncation and Filtering in Statistical Models Based on Counting Processes

摘要

The formulation of statistical models for life history data based on counting processes is reviewed. Examples include survival data with (possibly time-dependent) covariates and continuous time Markov processes. Complete life history data are discussed and explicit modeling of mechanisms causing incomplete information like (right) censoring, filtering, and left truncation is superimposed onto the model for the complete data. The concepts of independent censoring and noninformative censoring are defined, and inference from such models based on (partial) likelihoods is discussed.