NCJ Number
152065
Journal
Law and Human Behavior Volume: 18 Issue: 5 Dated: (October 1994) Pages: 587-593
Date Published
1994
Length
7 pages
Annotation
Researchers have offered several recommendations for describing the accuracy of violence prediction techniques using 2 x 2 contingency tables that treat the assessment of dangerousness as binary predictions about the future and portray prediction results as true positive, true negative, false positive, or false negative.
Abstract
Several problems are associated with the recommendations on describing clinical judgments about violence based on contingency tables. Such tables can be misleading because they conflate the intrinsic ability to detect future violence with risk level. Instead, the authors suggest that receiver operating characteristic (ROC) analysis be employed to quantify violence prediction accuracy. ROC analysis offers a way of portraying the accuracy of violence assessments so that clinicians can perceive the relationship between risk level and decision choice. 24 references, 2 tables, and 1 figure