Discrepancy Score Analysis: A Computer Simulation of Classification Stability

Abstract

Although the identification of a severe discrepancy between achievement and intellectual ability has been proposed as a diagnostic indicator of learning disabilities, questions regarding the technical adequacy of severe discrepancy score classification have not been carefully examined. This study simulated the effects of measurement error, instrument selection, and the cutoff standards used to define “severe” underachievement on the stability of severe discrepancy score classification decisions. A statistical software package was used to generate aptitude and achievement data, calculate regression-prediction discrepancy scores, and classify cases of severe underachievement for a hypothetical population of 5,000 cases. The results indicated that the actuarial classification of severe underachievement was disproportionately related to chance and instrument selection. The results are generalizable to any psychometrically-based classification procedure.