Advances in Spatial Statistical Methodology for Urban Crime

Statement of the Problem: To understand the dynamics of crime in urban areas, it is important to investigate the socioeconomic attributes of the communities as well as the interactions between neighborhoods. We contribute to two different areas of criminology literature, “crime and place” and “communities and crime,” through advances in two different classes of statistical methods: areal unit models and spatial point process models. Strong social ties between two neighborhoods means they are more likely to transfer ideas, customs, and behaviors. This implies that both crime and crime prevention could be transferred along these social ties. Social and geographic ties often have significantly different structures, motivating the need for a more flexible model for the dependence between communities. These methods could be used by practitioners to change the focus of interventions to social hubs and therefore improve the effectiveness of crime prevention policies. Subjects: Not applicable. Partnerships: We partner with the Social and Decision Analytics Division at the University of Virginia and the Arlington County Police Department, who provide local crime data and expertise of the Arlington County area. Research Design and Methods: We use call for service, use of force, and socioeconomic datasets to analyze the relationship between crime and explanatory factors. First, through new definitions of neighborhood matrices used in statistical models, we develop techniques and software to combine geographic and social proximity in spatial models for areal units to estimate domestic and sexual violence in Detroit, Michigan and Arlington, Virginia. We also create a framework for modeling areal unit covariates and crime point processes simultaneously, by addressing the problem of point-areal spatial misalignment. Through advances in the Log-Gaussian Cox process framework, we create novel contributions through a new model form and covariance structure for the relationship between violent crime and police use of force, while controlling for community demographics. Analysis: For the areal unit models, we test the new model against the existing state-of-the-art for five different types of spatial models and two different cities, so our analysis is reliable and robust. For the point process models, we also plan to test the modeling framework for multiple cities and model types. Products, Reports, and Data Archiving: The findings from this study will be published in a dissertation and will be published in peer-reviewed articles and presented at national and international conferences. We will also release open-source R software implementing the new modeling techniques with detailed user manuals.

Note: This project contains a research and/or development component, as defined in applicable law, and complies with Part 200 Uniform Requirements - 2 CFR 200.210(a)(14).

CA/NCF