Palm theory, random measures and Stein couplings (Q2075333)

The authors use Stein's method to prove a new explicit bound in Gaussian approximation in the Kolmogorov (or uniform) metric which is well-suited to being applied in conjunction with the tools of Palm theory or the theory of Stein couplings. Having established this bound, it is then used in numerous examples and applications. Applications when combined with Palm theory include Gaussian approximation for completely random measures (i.e., for which measures of each of a sequence of pairwise disjoint sets yield independent random variables), for the total excursion time of locally dependent random processes, and for the total edge length of Ginibre-Voronoi and Poisson-Voronoi tessellations. Stein couplings (of which the exchangeable pair and size-biased couplings are well-known examples) are used in conjunction with the main result of this paper to give Gaussian approximation bounds for sums of locally dependent random variables and for additive functionals in classical occupancy problems. Rates of convergence are shown to be optimal in many of the applications which are considered.