Logo image
Bayesian inference for spatially inhomogeneous pairwise interacting point processes
Journal article   Peer reviewed

Bayesian inference for spatially inhomogeneous pairwise interacting point processes

Matthew A Bognar
Computational statistics & data analysis, Vol.49(1), pp.1-18
04/15/2005
DOI: 10.1016/j.csda.2004.04.008

View Online

Abstract

Spatial point patterns are frequently modeled with pairwise interacting point processes. Unfortunately, inference in these models is complicated by the presence of an intractable function of the parameters in the likelihood. Because of the relative computational simplicity, frequentist inference in pairwise interacting point processes has dominated the literature. However, a Bayesian approach has not been computationally feasible until recently. Since the Metropolis–Hastings acceptance probability contains a ratio of two likelihoods evaluated at differing parameter values, the resulting intractable ratio complicates the required application of MCMC. In this article, we describe how to obtain Bayesian inferences without conditioning on the number of points in the pattern, allowing the modeling of spatial inhomogeneity in the density of points. After describing our importance sampling within MCMC algorithm, we analyze the well-known Irish drumlin data set using a hard-core Straussian model.
Metropolis–Hastings algorithm Spatial point pattern Gibbs point process Inhomogeneous pairwise interacting point process Bayesian inference Reversible jump MCMC Importance sampling

Details

Metrics

Logo image