|
Activity Number:
|
238
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #308457 |
|
Title:
|
Semiparametric Models in Bayesian Event History Analysis Using Beta Processes
|
|
Author(s):
|
Pierpaolo De Blasi*+
|
|
Companies:
|
University of Turin
|
|
Address:
|
Piazza Arbarello 8, Turin, 10122, Italy
|
|
Keywords:
|
Bayesian semiparametrics ; Bernstein-von Mises theorem ; beta processes ; competing risks model ; hazard regression ; Poisson random measures
|
|
Abstract:
|
We consider two topics of event history analysis: proportional hazards regression and competing risks models. Both are studied within new semiparametric formulations and the aim is to develop a full Bayesian treatment. The hazard regression model we consider is a variant of the Cox model, as it postulates a logistic relative risk function. The Bayesian construction involves a beta process for the cumulative baseline hazard and a Jeffreys-type density for the regression coefficients. The posterior distribution is derived and a Bernstein-von Mises (BvM) theorem is reached for our class of priors. Competing risks data describe failure times with multiple endpoints. We model cause-specific hazards (CSH) via the conditional probability of a failure type and the overall hazard. We propose a beta process for the overall cumulative hazard and derive a BvM theorem for the posterior of the CSHs.
|
