“…It is, however, relatively easy -though still computationally demanding and afflicted with substantial random error -to approximately compute expected values of empirical characteristics through Monte Carlo simulations by doing a random simulation of the process and computing characteristics from the simulation result. Such empirical characteristics can be any type of descriptive statistic, for instance, for a random closed set, the densities of the Minkowski functionals like volume fraction or specific surface, the covariance and certain contact distribution functions (Schneider and Weil, 2008;Ohser and Schladitz, 2009;Chiu et al, 2013), or, for a random point pattern, the intensity (function), the pair correlation function and the nearest neighbour distance distribution function (Møller and Waagepetersen, 2003;Illian et al, 2008), or any other sort of summary statistic maybe tailored specifically for the given estimation problem. In particular, this includes those cases where a spatial random structure is practically observable only via planar sections or projections and, hence, only statistics based on sections or projections are available.…”