|
Activity Number:
|
225
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 5, 2003 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Stat. Sciences
|
| Abstract - #300414 |
|
Title:
|
Semiparametric Bayesian Analysis of Matched Case-Control Studies with Missing Exposure
|
|
Author(s):
|
Malay Ghosh*+ and Bani K. Mallick and Samiran Sinha and Bhramar Mukherjee and Raymond J. Carroll
|
|
Companies:
|
University of Florida and Texas A&M University and University of Florida and University of Florida and Texas A&M University
|
|
Address:
|
223 Griffin-Floyd Hall, Gainesville, FL, 32611,
|
|
Keywords:
|
conditional inference ; Dirichlet process ; endometrial cancer ; equine epidemiology ; exponential family ; missing data
|
|
Abstract:
|
The paper considers Bayesian analysis of case-control problems with the exposure variable having a conditional distribution belonging to the exponential family and possibly containing some missing observations. A completely observed set of covariates may also be present along with the exposure variables. We consider a matched set-up where each stratum contains one case and M controls. The standard approach is to assume that the distribution of the exposure variable does not involve any isolated stratum effect except through the covariates. In contrast, we allow for the presence of stratum effect while modeling the distribution of the exposure variable. Consequently, for the retrospective conditional likelihood of the exposure variables, the stratum effects remain as nuisance parameters, which grow in direct proportion to the sample size. We assume a Dirichlet process prior with a mixing normal distribution for the stratum effects and estimate all the parameters in a Bayesian framework. The Bayesian procedure is implemented via Markov chain Monte Carlo numerical integration technique. Two matched case-control examples and a simulation study are considered.
|
