A Bayesian Model for Cluster Detection
Jonathan C. Wakefield, Albert Y. Kim
Departments of Statistics and Biostatistics, University of Washington, Seattle, WA, United States

An important problem in spatial epidemiology is the detection of areas in which the risk of a particular disease is significantly elevated, leading to an excess of cases. A number of frequentist approaches have been suggested for the detection of clusters, based on a hypothesis testing frame-work. However, these suffer from a number of drawbacks including the difficulty in specifying a p-value threshold at which to call significance, the inherent multiplicity problem, and the possibility of multiple clusters. In this talk we describe a Bayesian approach in which the study region is partitioned into, possibly multiple, “zones” within which the risk is either at a null, or non-null, level. Computation is carried out using Markov chain Monte Carlo, using an algorithm that is tuned to the cluster model. The latter is based on a scan statistic that underlies the populart SatScan methodology. The approach is applied to leukemia data in Upstate New York State.

Keywords: Bayes factors; Scan statistic; Spatial epidemiology; Markov chain Monte Carlo

Biography: Jon Wakefield has worked on population pharmacokinetics, spatial epidemiology, ecological inference, genetic epidemiology, and the analysis of data from high-throughput technologies, such as next generation sequencing. He has held positions in the Department of Mathematics, and the Department of Epidemiology and Public Health, at Imperial College. He moved to the University of Washington in 1999, where he holds a joint position in Statistics and Biostatistics. He is currently Chair of the Statistics Department.