|
Activity Number:
|
366
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #304493 |
|
Title:
|
Clipped Latent Variable Spatial Models for Ordered Categorical Data
|
|
Author(s):
|
Megan D. Higgs*+ and Jennifer Hoeting
|
|
Companies:
|
Montana State University and Colorado State University
|
|
Address:
|
2-214 Wilson Hall, Bozeman, MT, 59717-2400,
|
|
Keywords:
|
Bayesian ; stream health ; ordinal
|
|
Abstract:
|
Ecological and environmental research may produce ordered categorical data collected at point-referenced spatial locations. We combine and extend previously proposed methods to develop a Bayesian hierarchical model dealing with both the ordered categorical and spatial nature of such data. The approach relies on the use of a latent Gaussian random field and the notion of clipping the underlying continuous distribution to obtain a categorical random field. We discuss theoretical and practical issues involved in fitting the model, while focusing on prediction at new locations. We present simulation results, along with an application of the method to predict a surrogate measure of stream health in Montgomery County, Maryland.
|
