Sampling extreme groups invalidates research on the paraphilias: Implications for DSM-5 and sex offender risk assessments

摘要

Psychiatrist and Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV) text editor Michael First has criticized the addition of victim counts to criteria proposed by the Paraphilia Sub-Workgroup for inclusion in DSM-5 because they will increase false-positive diagnoses. Psychologist and Chair of the DSM-5 Paraphilia Sub-Workgroup, Ray Blanchard, responded by publishing a study of pedohebephiles and teleiophiles which seemed to show that victim counts could accurately identify pedohebephiles who were selected per self-report and phallometric testing. His analysis was flawed because it did not conform to conventional clinical practice and because he sampled groups at opposite ends of the clinical spectrum. In an analysis of his full sample, we found the false-positive rate for pedohebephilia at the recommended victim count selection points was indeed very large. Why? Because data analyses that eliminate intermediate data points will generate inflated estimates of correlation coefficients, base rates, and the discriminative capacity of predictor variables. This principle is also relevant for understanding the flaws in previous research that led Hanson and Bussiere to conclude that sexual recidivism was correlated with "sexual interest in children as measured by phallometric assessment." The credibility of mental health professionals rests on the reliability of their research. Conducting, publishing, and citing research that reflects adequate sampling and cautious diagnostic theorizing are critical for preserving this credibility.