Evaluating Classification Schemes Used in Reporting Criminal Justice Data

Three methodological topics are developed within the context of presenting statistical information through the use of some a priori classification schemes. First, a strategy is defined for evaluating the utility of a given classification that is based on, say, spatial contiguity or population. A heuristic optimization technique is then suggested as a way of constructing possibly better categorizations from secondary data sources. Finally, a method is given for comparing the adequacy of any two different classifications. Throughout the paper we discuss an example that involves crime rate data for 40 cities in the Los Angeles area