|
Activity Number:
|
139
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 4, 2008 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #301850 |
|
Title:
|
Mixed Model Selection Criteria Based on Statistical Computation and Leave-One-Out Method
|
|
Author(s):
|
Junfeng Shang*+
|
|
Companies:
|
Bowling Green State University
|
|
Address:
|
450 Math Science Building, Bowling Green , OH, 43403,
|
|
Keywords:
|
Akaike information criterion ; predictive divergence criterion ; Kullback-Leibler discrepancy ; improved Akaike information criterion
|
|
Abstract:
|
In the mixed modeling framework, statistical computation and leave-one-out method are employed to develop an improved Akaike information criterion and the predictive divergence criterion for model selection, respectively. The selection and the estimation performance of the criteria is investigated in a simulation study. The simulation results demonstrate that the predictive divergence criterion outperforms the improved Akaike information criterion in choosing an appropriate mixed model, and the improved Akaike information criterion is less biased than the predictive divergence criterion in estimating the Kullback-Leibler discrepancy between the true model and a fitted candidate model.
|
