|
Activity Number:
|
292
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #302860 |
|
Title:
|
Random Effect Seepage and Identification Problems in Bayesian Spatial Health Data Modeling
|
|
Author(s):
|
Andrew B. Lawson*+ and Sumirathan Rasathurai
|
|
Companies:
|
Medical University of South Carolina and University of South Carolina
|
|
Address:
|
135 Cannon Street, Charleston, SC, 29425,
|
|
Keywords:
|
random ; effect ; spatial ; seepage ; identification
|
|
Abstract:
|
Identification of random effects can be an important issue when models routinely employ a range of such effects. In spatial health modeling these effects can sometimes be grouped for estimation (such as in the classic convolution model). Identification from prior information is important in a Bayesian setting. What if we want to assess different types of identification in spatial models? For example we might want to consider additive random components or alternatively be concerned about the identification of random components when fixed effects are present (seepage). We will examine a range of scenarios where identification could be an issue. A summary measure will be considered that can be used as a global and local measure of the identification of both random effects and random and fixed effects. The extension to the selection of random and fixed effects will be considered.
|
