Mixture formulae for shot noise weighted point processes (Q1827548)

Gibbs or Markov point processes are widely used in spatial statistics as model for finite point patterns. For regular patterns, the family of Markov pairwise interaction models [see \textit{M. N. M. van Lieshout}, ``Markov point processes and their applications'' (2000; Zbl 0968.60005)] is particularly appealing. Members of this class are defined by a Radon-Nikodym density with respect to a Poisson process that can be written as a product over point pairs of some interaction function, typically specified in terms of the distance between the two points. For cluster patterns, Neyman-Scott or Cox models [see \textit{J. Møller} and \textit{R. Waagepetersen}, in: Spatial cluster modelling (2002; Zbl 1046.62102)] are natural candidates. In contrast to distance-based pairwise interaction functions, the Widom-Rowlinson style models focus on the influence of points on their environment. The same principle underlies the shot noise weighted point process models described by \textit{M. N. M. van Lieshout} and \textit{I. S. Molchanov} [Commun. Stat., Stochastic Models 14, 715--734 (1998; Zbl 0911.60025)]. Briefly, given a nonnegative influence function, the sum of influences over all points in a pattern is filtered by means of an interaction potential to yield the sufficient statistic of an exponential family. The authors prove that the components of any bivariate pairwise cross interaction point process form a shot noise weighted point process, thus extending results for the well-known Widom-Rowlinson penetrable spheres model. The authors also show that the first component of a bivariate shot noise weighted point process, with influence from points of the first component on those of the second one, is distributed as a shot noise weighted point process with the same influence function but a different potential. The paper is organized as follows. In Section 2 the authors review pairwise interaction and shot noise weighted point processes and fix notations. Sections 3 and 4 present the main results on the marginal distributions of multi-pairwise interaction and bivariate, shot noise weighted point processes, which are summarized in Section 5.