|
Activity Number:
|
180
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Risk Analysis
|
| Abstract - #302611 |
|
Title:
|
Model Averaging of Dosages from Toxicology Studies: Approaches and Pitfalls
|
|
Author(s):
|
Edward L. Boone*+
|
|
Companies:
|
Virginia Commonwealth University
|
|
Address:
|
Department of Statistics and Operations Research, Richmond , VA, 23284,
|
|
Keywords:
|
Dose Response ; Bayesian Methodology ; Model Averaging
|
|
Abstract:
|
Model averaging of dosages from toxicology studies have been recently introduced into the literature. When employing these averaging techniques often posterior model probabilities are used which are commonly approximated using BIC, KIC or some other information based criteria. While these approximations are valid asymptotically, there is no information on how accurate these methods are in low to moderate sample size cases. This presentation considers the performance of BIC and KIC against the fully Bayesian approach using posterior model probabilities generated by numerical integration. Examples of how to apply each approach and a comparison of each is considered.
|
