NCJ Number
92733
Date Published
Unknown
Length
41 pages
Annotation
Users of models of the criminal justice system should carefully analyze the potential performance measures when they choose a model, in order to ensure that the measurement process contained in the model and its accompanying methodology approaches the goal reflected in the measurement strategy.
Abstract
A model's performance measure, as well as the measurement strategy and process used, define the accuracy, reliability, and even the appropriateness of any estimate of behavior. Criminal justice researchers have used probabilistic models, nonlinear programing models, simulation models, and other models. In every case, the model's form and its accompanying methodology prescribe and restrict the measurement strategy or process and sometimes the type of performance measure that is appropriate. As a result, numerous assumptions and compromises are usually needed to make a model work. Since the choice of a performance measure for a model often imposes restrictions on the form of the model, model builders should consider the tradeoffs between performance measures and model forms. If they do this, they will be able to choose the approaches which best suit their needs. In simulations concerning career criminals and their effects on the criminal justice system, offenders may be portrayed either as individual entities or as aggregate flows. These two model forms can generate many measures of system performance. The individual offender simulation requires fewer assumptions for a greater number of these measures than does the flow model. A figure, data tables, and 23 references are provided.