Testing association between origin-destination spatial locations

Abstract

Point processes are probabilistic frameworks to analyze spatial patterns composed by random point features, called events, stored in GIS database. In a multivariate point process, the events are of two or more different types such as the locations of disease cases and a set of locations labelled as control individuals or as the positions of plants in a planar region labelled according to their species [2]. Usually, the spatial analysis of multivariate point processes is concerned with two questions. The first one concentrates on the comparison between the individual patterns of the component processes. Typically, the interest is to decide if one spatial pattern (such as the disease cases) has some degree of spatial clustering with respect to another spatial pattern (such as the controls’ pattern) (see [5]), perhaps identifying some putative sources of increased relative intensity ([3]; [4]). The second question concentrates on testing the independence of two (or more) point patterns and therefore attention is directed to the joint distribution of the processes [7]. It is common, for example, to test if the presence of an event of a certain type in a location either inhibits or stimulates the nearby presence of events of other types.