|
Activity Number:
|
218
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #305031 |
|
Title:
|
Bayesian Graphical Models for Multivariate Spatial Data with Application to Environmental Data
|
|
Author(s):
|
Kathryn M. Irvine*+ and Alix Gitelman
|
|
Companies:
|
Montana State University and Oregon State University
|
|
Address:
|
Department of Mathematical Sciences, Bozeman , MT, 59717,
|
|
Keywords:
|
Model Selection ; Multivariate Spatial Data ; Bayesian ; Graphical Models
|
|
Abstract:
|
A Bayesian graphical model provides a framework for visually displaying relationships between variables, and in turn the conditional and marginal independencies implied by the graph lead to simplifications of the entire multivariate joint distribution represented by the graph. Often this simplified probability distribution is relatively easier to model and interpret than the unsimplified version. We extend a particular graphical model---the isomorphic chain graphs (ICG) (Gitelman and Herlihy 2007)---to allow for multiple scales of spatial correlation. These models provide a flexible, more intuitive option for modeling multivariate spatial data because of the additional visual representation of the data. We present an environmental data set to illustrate the application of these models. Further we provide guidance on how to select among different graphical models based on simulations.
|
