Online Program Home
My Program

Abstract Details

Activity Number: 172
Type: Contributed
Date/Time: Monday, August 1, 2016 : 10:30 AM to 12:20 PM
Sponsor: Biometrics Section
Abstract #321284
Title: Semiparametric Bayesian Analysis of High-Dimensional Censored Outcome
Author(s): Chetkar Jha* and Yi Li and Steven Melly and Dr Subharup Guha
Companies: University of Missouri and University of Michigan and Harvard and University of Missouri
Keywords: CAR model ; Data Squashing ; Dirichlet Process ; Generalized Polya Urn Process ; Large Dataset ; Survival model

The SEER cancer database contains survival data for the U.S. population.The goal of this analysis is to discover spatial variation in breast cancer mortality rates in New Mexico while adjusting for known confounders such as race, age and tumor grade.

When analyzing large databases such as SEER using semiparametric Bayesian methods, Dirichlet process (DP) models exploits the richness of the dataset and provides information about breast cancer prognosis. But DP models can be prohibitively expensive for even a few hundred individuals.A cost effective MCMC strategy is applied to perform a fully Bayesian analysis of the SEER data.

The posterior distributions of several model parameters are highly non-normal. While a parametric model would make simplifying assumptions the semiparametric DP model flexibly adapts to arbitrary features of intersubject variation such as skewness and multimodality.There is strong evidence that after accounting for known indicators of disease prognosis, individual variability in breast cancer survival is non-normal and multimodal. This goes to show the value of DP mixture model and proposed fast MCMC algorithm.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2016 program

Copyright © American Statistical Association