357 – Bayesian Methods for Clinical Trials and Survival Analysis
Bayesian Hazard Change-Point Estimation with Incomplete Data
Deniz Yenigun
Istanbul Bilgi University
Ulku Gurler
Bilkent University
Hazard function is a fundamental tool in reliability and survival studies. Sometimes abrupt changes may be observed in the hazard function and there is a need for understanding the structure of this change. In this study we first focus on piecewise constant hazard functions with a single change-point when the observations are subject to truncation and censoring. From a Bayesian perspective, we discuss a method for estimating the time and size of the change. We then consider extending this work for piecewise linear hazard functions with a single change-point. The performance of the proposed estimation method is illustrated with a numerical study. Our results indicate that the Bayesian method performs well, and in some cases it may outperform the maximum likelihood method considered as a benchmark. An application to a real data set is also considered.