A Semiparametric Bayesian Approach For Analyzing Interval-censored Data

Interval-censored data are commonly encountered in survival analysis when the exact failure time is not available.Researchers can only observe a time interval that includes the failure time of interest.This paper focus on the general interval-censored data which is a mixture of left censoring,interval censoring and right censoring.Due to the complicated data structure,regression analysis for this kind of data is a challenging task.In this paper,based on the proportional odds model,we propose a novel semiparametric Bayesian regression analysis approach.In the analysis,Bernstein polynomials are adopted to approximate the baseline log odds function.Therefore,we can not only ensure the non-decreasing property of the baseline log odds function,but also obtain the estimates of parametric and non-parametric components simultaneously.Besides,by adopting the data augmentation method,the Gibbs sampler in this paper neither needs to impute any censored failure time,nor includes any complicated MetropolisHastings algorithms.Therefore,the algorithm is easy to implement.Finally,we conduct extensive simulation studies to evaluate the finite sample properties and apply the proposed approach to a real-life dataset.