Multiple State Models with Covariates: Analysis and Simulation.

摘要

The research reported in this document deals with a multiple state, staged failure phenonmenon with state specific covariates characteristic of the states. Different types of censoring was taken into consideration. The classical maximum likelihood estimation is compared with Cox's partial likelihood estimation for efficiency, since maximum likelihood estimation of covariate effects depends on the baseline failure rate, while partial likelihood estimation does not. Results of analysis as well as simulation are presented. (Author)