Robust Spatial Analysis of Rare Crimes

• NCJ Number: 205910
• Author(s): Avinash Singh Bhati
• Date Published: March 2004
• Length: 73 pages
• Grant Number(s): 2002-IJ-CX-0006
• Agencies: NIJ-Sponsored
• Publication Type: Report (Study/Research)

Annotation

This paper reports on a project that developed an analytical framework to be used for robust analysis of rare crimes typically observed at local (intra-city) levels of areal aggregation.

Abstract

As indicated in the first chapter, real-world problems such as discrete outcomes, finite samples, ill-conditioned data, spatial clustering, ill-measured regressors, etc., all preclude a simple adoption of the standard Ordinary Least Squares framework with its associated spatial-analytical toolkit. The second chapter of this report describes the semiparametric information-theoretic framework that achieves this goal. A second goal of this project was to examine and report on the extent to which structural and socioeconomic determinants of various kinds of violence (the disaggregated homicide types) may be distinct, as well as whether these findings persist at different (intra-city) levels of areal aggregation. The analysis was conducted on disaggregated homicide counts (1989-91) recorded in Chicago's census tracts and neighborhood clusters by using explanatory factors obtained from census sources. The research sought to determine how the socioeconomic and demographic characteristics in an area affect the amount of violence that community can expect to have, whether the links are violence-type-specific, and whether these links are areal aggregation level-specific. Given the explicit spatial nature of the data required to answer these questions, they must be examined in the presence of possibly spatially dependent errors. The information-theoretic approach developed in this project is used for this purpose. A final question addressed by this project is whether the inferences derived from the analysis would have been qualitatively different had the possible spatial structure in the errors been ignored and been qualitatively different had the possible spatial structure in the errors been ignored and traditional nonspatial modeling strategies been used. The data used to examine these issues are presented in the third chapter of this report, with findings discussed in the fourth chapter. The concluding chapter discusses the implications of this research and profiles promising extensions of the proposed analytical framework. The merits and limitations of the methodology are also addressed. 53 references