NCJ Number
85245
Date Published
1982
Length
409 pages
Annotation
The report develops procedures for comparing related structures represented as proximity matrices, concentrating on multidimensional scaling, clustering, and related data reduction methods, and includes papers on topics relating to epistemological and practical problems of aggregation, computer mapping methods, and the spatial effectiveness of past gun control laws.
Abstract
The strategy for evaluating data analysis schemes in order to clarify the structure underlying proximity measures is based on a technique for testing the difference between dependent correlations developed by Wolfe. A nonparametric procedure is used to assess the similarity in pattern between two proximity matrices. Discussion centers on possible uses of the method in analyzing criminal justice data. Numerical examples are given. An investigation of the association between State gun control laws and violent crime rates in 1975 shows that consideration of the data's geographic aspects can strongly affect the results of a correlation analysis between the two phenomena. A discussion of statistical graphic techniques indicates that unclassed statistical maps (those where data are not grouped into classes) are more accurate with regard to quantization error and appear to be of superior visual quality than maps with classes. A paper exploring epistemological and practical problems interfacing micro and macro approaches in geographical analysis conceptualizes them as a problem of hierarchical clustering. Papers include data tables and references.