Estimating operating characteristics from group mean difference research: a simple method for clinicians

UND Columbia Hall, 1350

Published research on psychological tests, including studies cited in some test manuals, frequently present statistically significant group mean differences (E.g., F or t statistics) as evidence of test validity and utility in clinical practice. Unfortunately, while such data may generally support a tests construct validity, they are useless to clinicians interested in data-driven decision making. Indicators of criterion-related validity, particularly operating characteristics (sensitivity, specificity, predictive power) are far more useful. We present a method for estimating operating characteristics from group mean difference designs, given certain assumptions. The method was recently tested using published data from such studies, and excellent prediction accuracy was found to be excellent. We will discuss the implications for clinicians, and issues not yet addressed by the estimation method.