Extended semiparametric mixture cure models for interval censored data

Analysis of the survival data with a subgroup of cured subjects arising in a clinical trial is commonly performed using a mixture cure model. Existing studies of the two-component mixture cure model assume a logistic regression for the cure probability and a conventional survival model for failure times of susceptible subjects. In this thesis, two extended semiparametric mixture cure models are proposed to analyze interval-censored data in which the failure times are recorded as intervals and there is a subgroup of subjects to be cured. The first proposal is to use the Bayesian doubly semiparametric mixture cure model to incorporate the nonlinear effects of risk factors both in the probability of being cured and the survival risks in the latency stage. The second proposal is based on a generalized accelerated hazards cure model to describe the time-scaled effects in the latency stage.Doctor of Philosoph