Testing local independence between two point processes.

Abstract

Dependencies between two types of points in a spatial point process can be due either to a real dependence between the two types or to the dependence on common underlying variables. We propose a global test for dependence between two point processes that is valid for a wide range of models. In contrast with previously proposed methods, it is based on a number of local test statistics, which makes it possible to map the local association between the two processes. The behavior of the test is evaluated by a simulation study. It is then applied to a vegetation pattern data set from Burkina Faso.