A0195
Title: Spatial scan statistics in statistical models
Authors: Tonglin Zhang - Purdue University (United States) [presenting]
Abstract: Spatial scan statistics are powerful in detecting spatial clusters for disease data, but the method is rarely combined with statistical models. To incorporate the method into a statistical model, a straightforward idea is to specify a group of artificial explanatory variables for spatial clusters with each cluster candidate explained by an explanatory variable. Based on this formulation, the spatial scan test can be carried out by a variable selection procedure with the estimates of coefficients for the strength for spatial clusters and the size of explanatory variables for the number of clusters. To implement this idea, the research develops a method to connect variable selection with cluster detection under the framework of generalized linear models. It then proposes a generalized information criterion approach to estimate both the number of clusters and their shapes.
