NCJ Number
186329
Journal
Criminal Justice and Behavior Volume: 27 Issue: 6 Dated: December 2000 Pages: 733-764
Date Published
December 2000
Length
32 pages
Annotation
This study used a large recidivism data set to compare the performance of a series of risk-screening devices that used the iterative classification procedure of Steadman and colleagues with performance of the more traditional linear and configural procedures.
Abstract
The iterative procedure started by separating cases into three initial categories on the basis of predicted risk and then reanalyzed cases in the intermediate risk category. The data came from a study by the New Jersey Administrative Office of the Courts and included information from official records of all offenders convicted of an indictable offense between October 1976 and November 1977. The research applied the iterative and traditional procedures across a range of recidivism outcomes and cross-validation conditions. The four recidivism measures included imprisonment within 1 year, imprisonment within 5 years, rearrest within 1 year, and rearrest within 5 years. Results revealed that the iterative classification procedure outperformed other standard device construction procedures in terms of the percentage of cases classified as high or low risk; however, the iterative classification did not outperform more traditional procedures on a variety of other performance measures. The analysis concluded that researchers and practitioners who are evaluating the performances of risk screening devices must be careful to select measures that correspond to the uses to which the devices would be put. Figures, tables, notes, and 28 references (Author abstract modified)