|
Activity Number:
|
186
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #301699 |
|
Title:
|
Hierarchical Model Selection Using a Benchmark Discrepancy
|
|
Author(s):
|
Lu Lu*+ and Michael D. Larsen
|
|
Companies:
|
Iowa State University and Iowa State University
|
|
Address:
|
2625 N. Loop Dr., Bldg. 2, Suite 2140, Ames, IA, 50011-1272,
|
|
Keywords:
|
generalized linear mixed model ; Poisson-gamma model ; Poisson-lognormal model ; posterior predictive checks ; small area estimation
|
|
Abstract:
|
In the context of small area estimation, hierarchical Bayesian (HB) models are often proposed to produce more reliable estimators of small area quantities than direct estimates. A method that benchmarks HB estimates with respect to higher level direct estimates and measures the relative inflation in the posterior mean square error of distributions due to benchmarking is developed to evaluate the performance of hierarchical models. The benchmarked HB posterior predictive model comparison method is shown to be able to select proper models effectively in a simulation study. The method is then applied to fitting models to a stratified multi-stage sample survey conducted by Iowa's State Board of Education. The survey strata serve as small areas for which HB estimators are suggested. Here the method is used to select a generalized linear mixed model for the survey data.
|
