|
Activity Number:
|
460
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics and the Environment
|
| Abstract - #305593 |
|
Title:
|
Clustering Using Stressor-Response Relationships
|
|
Author(s):
|
Samantha C. Prins*+ and Eric P. Smith
|
|
Companies:
|
Virginia Polytechnic Institute and State University and Virginia Polytechnic Institute and State University
|
|
Address:
|
Hutcheson Hall 0439, Blacksburg, VA, 24061,
|
|
Keywords:
|
stressor-response ; classification ; model comparison ; partition modeling ; Voronoi diagram ; model selection
|
|
Abstract:
|
Environmental data collected over large spatial regions often are analyzed using a single model though multiple models may be more appropriate. To find important models over smaller spatial regions, we will present a method for clustering sites to maximize the stressor-response relationships within clusters. The method uses Voronoi diagrams to assign sites randomly to one of K clusters based on values of nonstressor variables. Within each cluster, the method selects the optimal stressor-response relationship and fits the chosen model to all sites in that cluster. This form of model fitted is flexible and may be different across clusters. We will discuss methods to optimize the Voronoi assignment of sites to clusters, stressor-response model fitted within each cluster, and the number of clusters. We will present an application of the method to data from the mid-Atlantic Highlands.
|
