|
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 - #300216 |
|
Title:
|
Matched Case-Control Studies with Multiple Disease States and Multivariate Exposure: A Bayesian Approach
|
|
Author(s):
|
Bhramar Mukherjee*+ and Samiran Sinha and Malay Ghosh
|
|
Companies:
|
University of Florida and University of Florida and University of Florida
|
|
Address:
|
Statistics, Gainesville, FL, 32611,
|
|
Keywords:
|
conditional inference ; Dirichlet process ; Gibbs sampling ; missing exposure ; multiple disease state ; multivariate exposure
|
|
Abstract:
|
We consider two modeling issues in the context of a matched case-control study: (1) when there are multiple disease states, and (2) when the set of exposure variables under consideration are mutually associated. We develop a unified semiparametric Bayesian framework to approach these problems. For (1), we extend the classical multinomial logistic regression model to the analysis of matched data. For (2), we model the underlying association scheme when the exposure is multivariate binary or each exposure is measured on a continuous scale, and also for the "mixed" case when some are binary, some are continuous. The method we propose can handle possible missingness in the exposure variables and also allows for the possible presence of stratum effect on the distribution of the exposure variables. The Bayesian estimation procedure is implemented via the Markov chain Monte Carlo numerical integration technique. Examples of real case-control studies are used to illustrate the methodology.
|
