Testing Complete Spatial Randomness on Linear Networks: Leon County Traffic Accident Example

Abstract

A relatively new sub-area within this is the statistical analysis of point processes on linear networks, that is, processes of events occurring randomly in space but with locations constrained to lie on a linear network. For example, traffic accidents occur at random locations constrained to lie on a network of streets. In this case, the network is idealized as a network of line segments in the plane or three-dimensional space. The development of statistical techniques for the analysis of point processes on linear networks is still in its infancy. Many standard statistical techniques for analyzing point processes cannot be directly applied to data arising from linear networks and require suitable modification. Test of Complete Spatial Randomness (CSR) for point processes on the plane based on quadrat counts or nearest neighbors cannot be applied to point processes on linear networks. This paper defines a Voronoi tessellation of the linear network which uses the shortest path distance along the network instead of Euclidean distance, and then develops a chi-square test of CSR for linear networks based on the event counts in the tiles of this tessellation. This test is applied to data on traffic accidents in Leon County, Florida, USA.