Latent Process Segmentation for Point Process Data

• Activity Number: 335
• Type: Topic Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): David Kessler*+ and David Dunson and Brian Reich
• Companies: The University of North Carolina at Chapel Hill and Duke University and North Carolina State University
• Keywords: point process ; latent process ; geostatistics ; gaussian process

Abstract

Models for geostatistical and other point process data must be general enough to describe natural processes but simple enough to allow tractable computation. We will develop a class of flexible models for these data that employ latent processes to segment the spatial domain. Under this treatment, the elements of the segmentation are described by homogeneous models, allowing complex phenomena to be described by principled combination of straightforward component processes. The theoretical properties of the segmentation technique, including the demonstration of large support on a broad class of segmentations are described and the Bayesian framework that develops these models is introduced. General applicability of the approach is illustrated through application to data sets from different disciplines.