|
Activity Number:
|
171
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #304900 |
|
Title:
|
Generalized Additive Models with Spatio-Temporal Data
|
|
Author(s):
|
Xiangming Fang*+ and Kung-Sik Chan
|
|
Companies:
|
East Carolina University and The University of Iowa
|
|
Address:
|
Department of Biostatistics, Greenville, NC, 27858,
|
|
Keywords:
|
Generalized Additive Model ; Matérn Model ; Penalized Likelihood ; ML ; REML ; Spatio-Temporal Data
|
|
Abstract:
|
Generalized additive models (GAMs) have been widely used. While the procedure of fitting a GAM to independent data has been well established, the available fitting approaches for correlated data are often numerically unstable or computationally intensive. In this paper, we propose a new iterative algorithm for penalized ML and REML estimation in fitting a GAM with correlated errors. The Matérn model is investigated for the case of separable spatio-temporal data with fixed spatial covariate structure and no temporal dependence. The conditions of asymptotic posterior normality are discussed. We also propose a new model selection criterion for comparing models with and without spatial correlation, and a model diagnosis method for checking the assumption of temporal independence for spatio-temporal data. The proposed methods are illustrated by simulation studies and a fisheries application.
|
