Spatial Model Selection

• Activity Number: 183
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Ciro Velasco-Cruz*+
• Companies: Virginia Tech
• Keywords: Model Selection ; Spatial Analysis ; Neighbor System

Abstract

Model selection combined with Spatial Analysis lead us to an interesting problem: how to select the best model(s) within subregions of a region of study. This will raise several difficulties: Find the best way to split the region into subregions that could make sense for further analysis and also for its interpretation, and Select a model taking into account the dependence that those subregions would entail. It is assumed that the presence of a given regressor into the selected model within a subregion would conform to the dependence structure in the whole region. It would mean that if a particular regressor is important in the selected model in a subregion but not in any other sub-region across the region, this would be because in that region a particular event is taking place that causes the regressor to be important, because the cause-effect its neighboring regions are inducing.