Survival Analysis, Nonparametric

Abstract

Abstract Survival analysis was one of the first areas where Bayesian nonparametric priors were utilized to enhance the flexibility of standard parametric models. Bayesian nonparametric methods are very well suited for survival data analysis, enabling flexible modeling and rich inference for the survival function or hazard function, providing techniques to handle censoring, and allowing incorporation of prior information. In this article, we provide a review of Bayesian nonparametric and semiparametric approaches, based on mixture models, to the analysis of univariate survival data. We discuss mixture models for survival distributions in a fully nonparametric setting as well as for error distributions in semiparametric accelerated failure time regression settings.