NCJ Number
211347
Journal
Criminal Justice and Behavior Volume: 32 Issue: 5 Dated: October 2005 Pages: 565-585
Date Published
October 2005
Length
21 pages
Annotation
This study examined the generalizability of the Level of Service Inventory-Revised (LSI-R) and the Violence Risk Appraisal Guide (VRAG) probability bins in a predominantly violent correctional sample.
Abstract
In recent years, many instruments have been introduced as measures of criminal and violence risk prediction. Two instruments are the Level of Service Inventory-Revised (LSI-R) and the Violence Risk Appraisal Guide (VRAG) which were developed from different theoretical perspectives, on different samples, and to predict different recidivism outcomes. This study examined generalizability of the probabilities associated with the original bin structure of the LSI-R and VRAG risk prediction instruments in a sample different from the initial validation study and offered an empirical approach to establishing a bin structure. The suggested empirical approach to binning focused on possible three-bin solutions for both instruments. The study consisted of 209 volunteers drawn from a population of incarcerated adult men sentenced to 2 years or more and who participated in the psychological assessment process. The results do not support the generalizability of the original probabilities associated with the prediction bins, although the LSI-R bins performed much better than the VRAG bins. Overall, the original LSI-R probabilities tended to underestimate the likelihood of general reoffending and the original VRAG probabilities tended to overestimate the likelihood of violent reoffending. The study serves to introduce the need for closer examination of probability bins in the prediction of recidivism in order to determine the statistically preferred approach that will identify linear bins with minimal score distribution. Figures, tables, and references