ABSTRACT
Advances in computing and Geographic Information Systems (GIS) mean that more scientific approaches can now be utilised to understand patterns of crime and their distribution more effectively. These include statistical and mathematical techniques, such as ‘Kernel Density Estimation’, which can be used to explore crime density and highlight ‘hotspots’ of crime. However, these techniques are predominantly descriptive in their approach. Approaches such as ‘regression’ offer a clear advantage over those that can only describe or illustrate patterns of crime, in that they seek to explore the impact of explanatory variables (Chainey and Ratcliffe, 2005). However, Chainey and Ratcliffe (2005) note how standard
linear regression models are problematic when focusing on geographical spatial data because variables are assumed to exert equal influence across an area despite the fact that this is unlikely to be the case. Furthermore, while methods such as geographically weighted regression account for the variance in independent variables across spatial areas, such methods still seek to account for crime at the aggregate rather than individual-level. Therefore, the richness of data at the individual level is lost using this approach.
