skip to primary navigation skip to content

Department of Geography

 

Bayesian spatial modelling of police-defined high-intensity crime areas – a case study of Sheffield

Research Team: Robert Haining and Jane Law

Abstract

The research models the geographical distribution of high crime areas using a dataset of the city of Sheffield in England. Earlier work, now published (Craglia et al 2001) reported the findings of an analysis of police-defined high intensity crime areas (HIAs) for a sample of police force areas in England. This work is now being extended with a more detailed examination of crime patterns in Sheffield using crime statistics (by small area) and the Sheffield police’s own definition of where their HIAs are located. The research reviews the earlier work and then models the Sheffield dataset using a Bayesian approach. The basic model is the logistic regression (an enumeration district is either in an HIA or it is not). The Bayesian approach extends the standard logistic regression model by including the spatially structured and/or unstructured random effects. The spatial (conditional autoregressive and intrinsic conditional autoregressive) and non-spatial models were assessed using the Deviance Information Criterion and Misclassification Statistics (proportion of enumeration districts belonging to police-defined HIAs that were not classified as such by the models). The results reveal that the Bayesian approach that uses a model with the spatially structured and unstructured random effects is better in modelling the police-defined HIAs. To visualize the results of this model, ‘map decomposition’ is used to identify high or low risk areas and the extent to which each of these risk areas are driven by the covariate or spatially structured random effects. Finally, to provide more information on the relationship between each enumeration district and whether it is in an HIA or not, the study compares two ways to map the estimated probabilities of enumeration districts in HIAs.

Illustrations

Mapping the estimated probabilities of enumeration districts in HIAs

Image as described adjacent

Further reading

  • Law, J., and R. P. Haining (2004). “A Bayesian Approach to Modeling Binary Data: The Case of High-Intensity Crime Areas.” Geographical Analysis, Vol. 36 (3), p. 197-216.
  • Craglia, M., R. Haining, and P. Signoretta (2001). “Modelling High-Intensity Crime Areas in English Cities.” Urban Studies 38(11), 1921-1941.