Department of Statistical Science
Duke University
presents:
Robert L. Wolpert
Duke Statistics, Duke University
"Spatial Modeling with Inhomogeneous Point-process Random Fields:
Bayesian Models for studying Disease Patterns and Biodiversity"
Abstract: We construct and study hierarchical spatial point process models, using inhomogeneous Poisson random fields whose random intensity measures are mixtures of gamma, stable, or other infinitely-divisible fields. This class of models is conjugate for local census observations in which the locations of all points within some subregion are recorded. Unlike the usual gamma/Poisson models, these feature stochastically dependent intensity measures at different sites and can be used to express prior beliefs about homogeneity, continuity, and similar features for the intensity. Bayesian posterior and predictive distributions are computed through the use of Markov chain Monte Carlo computational methods.
The models and techniques are then applied to the analysis of spatial dependence in reported incidence of Sudden Infant Death Syndrome (SIDS) in North Carolina and to the problem of assessing and quantifying regional patterns in biodiversity, investigating the density variation of each of several species of overstory tree within a research tract in Duke Forest.
October 11, 1996
4:00 pm - 5:00 pm
116 Old Chem Building Any questions concerning the seminar may be addressed to Cheryl McGhee @[919] 684-8029, e-mail cheryl@stat.duke.edu