ABSTRACT
The current practice for the analysis of spatial distributions of road collisions relies almost entirely upon visually examining a map showing the location of collisions, superimposed upon the road network. The assessment process is subjective and relies heavily on exercising judgment in order to decide whether there is a distinct pattern and what it is. Over the years, there has been an increasing move to study road collisions as cluster events. This in itself has brought a number of advantages and subsequent challenges to spatial analysis. Let us, first of all, take the example of crime. Crime does not occur randomly. There is a clear spatial dependence; and spatial heterogeneity exists when analyzing crime. Crime is affected by the surrounding area, and Tobler’s first law of geography can be applied here: “Everything is related to everything else, but near things are more related than distant things” (Tobler 1970, 236). Crimes, unlike road collisions, are not constrained to the road network, and therefore the clustering of events arguably is somewhat different.
