Evaluation of Software for Displaying and Analyzing Crime Patterns and Trends: Results From Five GIS and Spatial Statistical Software Analyses of Murder Data From the Bronx, New York

• NCJ Number: 173181
• Author(s): D Williamson; V Goldsmith
• Date Published: 1997
• Length: 28 pages
• Grant Number(s): 95-IJ-CX-0103
• Sponsoring Agency: National Institute of Justice (NIJ); US Dept of Justice NIJ Pub
• Sale Source: National Institute of Justice/NCJRS; NCJRS Photocopy Services
• Language: English
• Country: United States of America

Annotation

This study evaluated two GIS (Geographic Information Systems) software packages for displaying and analyzing crime patterns and trends, as well as three spatial statistical software packages.

Abstract

The two GIS packages evaluated were ArcView v3.0 and MapInfo v4.0. The compatibility of data between these packages is not straightforward, sometimes requiring the purchase of extra programs, and not even then in some cases; however, once police data exist in a database such as FoxPro, Oracle, dBase, Excel, etc., they can usually be incorporated within any of these GISs. For mapping police data by using automatic address matching (i.e., geocoding), similar rates of only 60 percent of the total were achieved by using ArcView and MapInfo, largely due to the inconsistencies in the formatting of either the crime data or in the local address database derived from the Census TIGER files. Once the data were incorporated within the GIS, comparison of the resulting maps with other mapped geographic data, such as comparison of murder rates with income data, proved to be relatively easy to perform. The three spatial statistical software packages evaluated are Stat!, Cluster, and S-Plus, which are among the few statistical packages that explicitly include spatial analyses. Whereas Stat! and especially S-Plus provide tools for Exploratory Data Analysis, Stat! provides more useful "true" statistical analyses. Data input is a major problem in these packages because of their strict requirements for formatting the data. More often than not, the data had to be "massaged" in order for them to have a compatible input format. Specifically, "time" parameters need special attention. Once the data are successfully input, the process is relatively quick and simple in all three software packages; however, one problem with these statistical tests is that the user must be familiar with the data and with the tests, as well as understand the specific statistical theory and its strengths and weaknesses sufficiently to interpret the results. 18 figures and 10 references