Evaluation of Classification Error in a Survey on Sexual Assault Among College Students
Marcus Berzofsky
RTI International
Chris Krebs
RTI International
Christine Lindquist
RTI International
Michael Planty
RTI International
Lynn Langton
Bureau of Justice Statistics
This paper evaluates the level of classification error that may exist when surveying college students regarding sexual assault. How sexual violence is defined within a survey can impact how a student responds to the victimization prevalence items. If a student misunderstands the meaning of the definition or does not feel the definition fits their exact circumstances, the result may be classification error - a student misclassifying their true sexual assault status. To address this concern, the Campus Climate Survey and Validation Study (CCSVS), conducted in the spring of 2015 at nine post-secondary institutions, incorporated multiple indicators of sexual assault. Using these multiple indicators, this paper shows the results of a latent class analysis (LCA). LCA estimates the false positive - an indication a person was victim when their true status is not a victim - and a false negative rate - an indication a person is not a victim when their true status is a victim. Our findings show that a false negative classification error is larger problem than a false positive. The reported victimization rates in the CCSVS are likely slight underestimations of the true victimization rate.