Weak and strong convergence of empirical distribution functions from germ-grain processes

Lothar Heinrich, Zbyněk Pawlas

Statistics · 2008 · 13 citations · 15 references

Concepts

Abstract

We observe randomly placed random compact sets (called grains or particles) in a bounded, convex sampling window W n of the d-dimensional Euclidean space which is assumed to expand unboundedly in all directions as n→∞. In addition, we suppose that the grains are independent copies of a so-called typical grain Ξ0, which are shifted by the atoms of a homogeneous point process Ψ in such a way that each individual grain lying within W n can be observed. We define an appropriate estimator ˆ F n (t) for the distribution function F(t) of some m-dimensional vector f(Ξ0)=(f 1(Ξ0), ˙s, f m (Ξ0)) (describing shape and size of Ξ0) on the basis of the corresponding data vectors of those grains which are completely observable in W n . As main results, we prove a Glivenko-type theorem for ˆ F n (t) and the weak convergence of the multivariate empirical processes √Ψ(W n )(ˆ F n (t)−F(t)) to an m-dimensional Brownian bridge process as n→∞.

References

15