B1107
Title: Bayesian spatial and spatiotemporal cluster Regression modeling
Authors: Andrew Lawson - Medical University of South Carolina (United States) [presenting]
Abstract: Bayesian cluster modeling can be approached in a variety of ways. Explicit parametric cluster models can be assumed, or as an alternative, exceedance probabilities can be used, which are associated with an underlying standard risk model. These are often limited in their ability to link predictors to clustering effects directly. We will examine the use of mixture models whereby the spatial or spatio-temporal risk surface consists of a mixture of uncorrelated and correlated effects. The mixing probability is allowed to be spatially or spatio-temporally dependent, and is linked to a linear or non-linear predictor via a standard link function (logit, probit etc.). An example of cluster modeling for respiratory cancer incidence at a local spatial scale will be provided.
