Abstract
In a previous issue of the Journal of Threat Assessment, an article by Wollert (2002b) documented an analysis of the actuarial risk assessment instrument known as the Minnesota Sex Offender Screening Tool-Revised (MnSOST-R). Based on that analysis, he drew the conclusion that a cross-validation study of the MnSOST-R showed a very significant degree of “shrinkage” in the percentages of recidivists associated with scale score risk categories. Additional conclusions indicated that the MnSOST-R was not a proper instrument for use within the sex offender civil commitment context. This article takes issue with the underlying analysis conducted in Wollert. The analysis documented herein serves to dispute the conclusion drawn concerning cross-validation shrinkage in the MnSOST-R's recidivism risk percentages. Essentially, the same type of comparison used by Wollert (2002b) was employed, but using more proper numbers. The new results indicate that, in fact, there was no shrinkage in risk percentages across the scale score categories. The effects of this finding on Wollert's (2002b) ultimate conclusions are discussed, along with general recommendations concerning use of the MnSOST-R and response comments to Wollert's (2002a) commentary to this article.