Citation

Wollert R. J. Threat Assess. 2002; 2(1): 87-102.

Copyright

(Copyright © 2002, Informa - Taylor and Francis Group)

DOI

10.1300/J177v02n01_06

PMID

unavailable

Abstract

Many states have passed laws for the civil commitment of sexually violent predators. Experts who testify at commitment hearings often base their opinions on actuarial tests for the prediction of sexual recidivism (ATSRs). Reliance on actuarial tests is promising, but they need to be accurate. Cross-validation is a widely recommended procedure for determining accuracy. In this paper the predictive accuracy of an ATSR called the Minnesota Sex Offender Screening Tool-Revised (MnSOST-R) was evaluated from a cross-validational perspective. The original predictions of the MnSOST-R were found to be greatly inflated. Dissemination of the MnSOST-R without adequate cross-validational research raises serious doubts about its adequacy. It also creates doubt in the professional community that other ATSRs might have been disseminated without sufficient attention to important test construction standards. Formulating a set of standards targeted specifically on the construction of ATSRs would be helpful to test developers, test users, and the court for evaluating their adequacy. Such a set of standards is proposed by the author. It is hoped that they will promote an adjustment of values so that test construction principles will be accorded the importance they deserve as ATSRs are developed in the future.