Predicting school weapon possession: A secondary analysis of the youth risk behavior surveillance survey



As school psychologists are asked to help schools evaluate the level of violence risk posed by specific students, they need to fully consider the technical adequacy of any test or procedure that is proposed to “predict” future violent behavior or conditions. In this article, we contribute to this discussion by examining responses of 40,435 students from the 1993, 1995, and 1997 administrations of the Youth Risk Behavior Surveillance Survey. Student self-reported school risk behaviors and experiences are used to “predict” recent weapon possession on school campuses and to illustrate the use of receiver operating characteristic curves to evaluate the performance of a test for prediction purposes. An index of nine school risk behaviors was moderately correlated (r = .36) with school weapon possession. However, when the accuracy of using this school risk index to predict weapon possession is evaluated, it is found that the area under the ROC curve was approximately .75, a moderate-to-large effect size but still demonstrating inadequate prediction at the individual student level. Further, there were twice as many frequent weapon carriers with zero school risks than with 7–9 risks, suggesting that attention directed toward violence risk profiles may promote inattention to another group of students who engage in potentially high-risk behaviors at school. It is argued that school psychologists should consider school violence appraisals within a developmental, long-term model. © 2001 John Wiley & Sons, Inc.