|
Activity Number:
|
348
|
|
Type:
|
Luncheons
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 7:00 AM to 8:15 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #306994 |
|
Title:
|
Model Selection in Hierarchical Models
|
|
Author(s):
|
David B. Dunson*+
|
|
Companies:
|
National Institute of Environmental Health Sciences
|
|
Address:
|
NIEHS, Biostatistics Branch, Research Triangle Park, NC, 27709,
|
|
Keywords:
|
model selection ; random effects ; variance components ; homogeneity test ; Bayes factor ; BIC
|
|
Abstract:
|
Traditional methods of model selection, such as likelihood ratio tests and AIC/BIC selection criteria, may not be appropriate for comparing hierarchical models with different covariance structures or numbers of latent variables. Some of the difficulties arise due to constraints on the parameter space. We will discuss emerging methods for dealing with model uncertainty in hierarchical models, such as random effects models and structural equation models. Some possibilities include new approximations to the Bayes factor and stochastic search methods. Important issues include prior selection and methods for efficient computation.
|
