A0943
Title: Semiparametric Bayesian discrete event time modeling
Authors: Adam King - California State Polytechnic University, Pomona (United States) [presenting]
Abstract: Many event time outcomes are discrete, either because the process is inherently discrete (e.g., number of semesters to graduation) or because an underlying continuous process has been discretized (e.g., time to cessation of drug use recorded as whole number of months). Existing approaches for handling such discrete data include treating the survival times as continuous (with adjustments for tied outcomes), or using discrete models that omit important features like random effects. We present a general Bayesian discrete-time proportional hazards model, incorporating a number of features popular in continuous-time models such as competing risks, frailties, and generalized additive models style semiparametric incorporation of time-varying covariates (including flexible baseline hazards for time effects). These methods are implemented in a freely available R package called brea. We illustrate with analyses of college graduation rates and time to illicit drug use cessation.
