Perfect sampling for posterior landmark distributions
with an application to the detection of disease clusters

Marc A. Loizeaux and Ian W. McKeague

Abstract:

We study perfect sampling for the posterior 
distribution in a class of spatial point process models
introduced by Baddeley and van Lieshout (1993).
For Neyman--Scott cluster models,
perfect sampling from the posterior is shown to be computationally 
feasible via a coupling-from-the-past 
type algorithm of Kendall and Moller.    
An application to data on leukemia incidence in upstate 
New York is presented.