Regularity conditions in the realisability problem with applications to point processes and random closed sets (Q2258527): Difference between revisions

Defining the distribution of a random element \(\xi\) in a topological space \(\mathcal X\) is equivalent to specialising the expected values for all bounded continuous functions \(g(\xi)\). It is well known that a linear functional \(\Phi(g)=\operatorname{E}g(\xi)\) on the space of bounded continuous functions \(g: {\mathcal X}\mapsto {\mathbb R}\) corresponds to a random element if and only if \(\Phi\) is positive and upper semi-continuous (\(\operatorname{E}\) stands for expectation). The authors consider the case of a functional \(\Phi\) defined only on some functions on \(\mathcal X\) and address the realisability of \(\Phi\), that is, the mere existence of a random element \(\xi\) such that \(\Phi(g)=\operatorname{E}g(\xi)\) for \(g\) from the chosen family of \(G\) of functions. This paper focuses on some geometric instances of the problem. Moreover, the answer of the existence problem consists of the two steps. The first step (positivity) ensures that it is possible to extend the functional \(\Phi\) positively from a certain family of functions to a wider family and the second step checks the regularity condition. The use of positive extension techniques in the framework of stochastic geometry was pioneered in [\textit{T. Kuna} et al., Ann. Appl. Probab. 21, No. 4, 1253--1281 (2011; Zbl 1235.60052)] for point processes. In this paper, the authors establish the general nature of an idea proposed in [loc. cit.] and show how it leads to various further realisability results. The new idea is to introduce an additional function -- regularity modulus -- and to formulate sufficient and necessary conditions in terms of a positive extension of a functional onto the linear space containing the regularity modulus, and requiring only a priori integrability of the regularity modulus. The authors concentrate on two basic examples of the realisability problem: the existence of point processes with given correlation (factorial moment) measure and the existence of a random closed set with given two-point covering probabilities or contact distribution functions. The introduction to the realisability issue for point processes is available in [loc. cit.] and [\textit{T. Kuna} et al., J. Stat. Phys. 129, No. 3, 417--439 (2007; Zbl 1130.60056)]. The realisability problem for random closed sets has been widely studied in physics and material science literature, and in particular the comprehensive monograph [\textit{S. Torquato}, Random heterogeneous materials. Microstructure and macroscopic properties. New York, NY: Springer (2002; Zbl 0988.74001)] and a recent survey [\textit{J. A. Quintanilla}, Proc. R. Soc. Lond., Ser. A, Math. Phys. Eng. Sci. 464, No. 2095, 1761--1779 (2008; Zbl 1167.15015)]. The paper is organized as follows. Section 2 presents a series of general results on regular extensions and also invariant extensions. Section 3 presents a number of realisability conditions for correlation measures of point processes that considerably extend the results of [Kuna et al., loc. cit.] by relaxing the moment and hard-core conditions. Section 4 deals with the realisability problem for two-point covering probabilities of random sets. The closedness of the corresponding random set can be ensured by imposing appropriate regularity conditions. Section 5 presents a further variant of the realisability problem that involves contact distribution functions of random sets.