Analysis of Progressively Censored Competing Risks Data

Publisher This chapter develops inference for a competing-risk model under a very general censoring scheme. Censoring is inevitable in life-testing and reliability studies because the experimenter is unable to obtain complete information on lifetimes for all individuals. For example, patients in a clinical trial may withdraw from the study, or the study may have to be terminated at a pre-fixed timepoint. In industrial experiments, units may break accidentally. In many situations, however, the removal of units prior to failure is pre-planned to provide savings in terms of time and cost associated with testing. The two most common censoring schemes are termed “type I” and “type II” censoring. The chapter considers competing risk data under progressive type II censoring and focuses on the analysis of the competing risk model when the data are progressively type II censored. The estimation of the different parameters is considered and their distributions are presented. Bayesian analysis of the competing-risk model is discussed in the chapter. Results of a simulation study comparing the coverage probabilities and lengths of the different confidence and credible intervals are also provided and the performance of these different techniques using a real dataset are illustrated.