MPS-RR 2001-5
February 2001

A comparison of spatial point process models in epidemiological applications

Jesper Møller

Abstract

As pointed out in Richardson (2001) (for short, "SR"), point process models provide a natural setting in epidemiological applications. The aims of this contribution are to review the advantages and problems when using such models for aggregated data, compare the properties of different model classes, clarify some of the properties of these model classes in connection to epidemiological applications (which sometimes seem to have been misunderstood), propose some alternative models which are not covered in SR, and point out some open problems. The focus is on the particular model classes: the Heikkinen and Arjas (1998) models, log Gaussian Cox (LGC) processes (Møller et. al, 1998), and shot-noise G Cox (SNGC) processes (Brix, 1999). Poisson/gamma models (Daley and Vere-Jones, 1988; Wolpert and Ickstadt, 1998) are special cases of SNGC processes.

Availability: [ gzipped ps-file ] [ pdf-file ]

[ Help on down-loading/viewing/printing ]