Nonparametric Bayes Estimators Based on Beta Processes in Models for Life History Data

摘要: Several authors have constructed nonparametric Bayes estimators for a cumulative distribution function based on (possibly right-censored) data. The prior distributions have, example, been Dirichlet processes or, more generally, neutral to the right. present article studies related problem of finding hazard rates and quantities, w.r.t. that correspond rate with nonnegative independent increments. A particular class processes, termed beta is introduced shown constitute conjugate class. To arrive at these, time-discrete framework survival data, which has some interest, studied first. An important bonus approach hazards complicated models life history data than simple table situation can be treated, time-inhomogeneous Markov chains. We find posterior derive in such also semiparametric Bayesian analysis Cox regression model. are easy interpret compute. In limiting case vague solution Nelson-Aalen estimator probability Kaplan-Meier estimator.