Citation

Mills JF, Jones MN, Kroner DG. Crim. Justice Behav. 2005; 32(5): 565-585.

Copyright

(Copyright © 2005, SAGE Publishing)

DOI

10.1177/0093854805278417

PMID

unavailable

Abstract

Statistics such as the Pearson's r and Receiver Operating Characteristics are often used to test the generalization of criminal and violence prediction instruments. However, these analyses overlook potential error in the assessment of risk if the rates of offending within the initial validation samples are assumed accurate for other samples. This study examined the generalizability of the Level of Service Inventory-Revised and Violence Risk Appraisal Guide probability bins in a predominantly violent correctional sample. The findings showed that the initial bin probabilities were not transferable to our sample of predominantly violent male offenders. An empirical method of optimal binning was introduced. The discussion centered on the accurate use of bin probabilities in the communication of risk.

Language: en