Analysis of interval-censored data: A simulation study

摘要

In many studies in which the response variable is the time until the occurrence of an event, the exact time cannot be determined, that is, only the interval of the occurrence is known. Such data can be analyzed by the traditional life table method (LTM) when there is no covariate. A more general approach consists in using a discrete-time regression model, such as proportional hazard model (DCM) or proportional odds model (DLM). In this paper we compare those three types of analysis (LTM, DCM, DLM) for the two-sample case through a simulation study. We assess the agreement among them with respect to the comparison between two groups, as well as the empirical power and the length of the confidence interval for quantities of interest. We also investigate the impact of the misspecification of the regression model.