|
Activity Number:
|
345
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 8, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #307273 |
|
Title:
|
Data Augmentation within a Conditionally Specified Gaussian Spatial Model
|
|
Author(s):
|
Brooke Fridley*+ and Philip Dixon
|
|
Companies:
|
Mayo Clinic College of Medicine and Iowa State University
|
|
Address:
|
200 First Street, SW, Rochester, MN, 55905,
|
|
Keywords:
|
censored ; data augmentation ; spatial ; Bayesian
|
|
Abstract:
|
Censored data occurs in numerous areas of application. When one adds the complexity of spatial dependency between observations, methods for handling censored observations become numerically challenging. Often in environmental studies, censoring occurs when contamination values fall below a level of detection. A common practice for handling censored observations is to set the censored observations equal to a function of the level of detection. Instead of using this single imputation approach for censored observations, data augmentation for the censored observations can be implemented. The use of data augmentation within a Bayesian conditionally specified Gaussian spatial model will be illustrated. In doing so, results from an analysis of a dioxin contamination site and results from a simulate study will be presented.
|
