|
Abstract:
|
Traditional approaches to modeling spatial clusters typically focus on localized clusters, but may not adequately account for clustering patterns exhibited globally across the region of interest. This presentation will describe a spatial modeling approach that starts with the full region, and uses ideas from functional data analysis to provide a global clustering concept. The proposed approach will be shown to increase power over traditional methods when there is a possibility of multiple clusters, but still retain very good power in instances of a single isolated cluster. The technique is furthermore coupled with classical sequential process-control methods to produce a very flexible spatio-temporal monitoring method. Demonstrations are presented in a disease-rate monitoring context, in which the data arrive as monthly, county-by-county disease counts across a state.
|
