A COMPARISON OF THE IMPACT OF REGRESSION AND SIMPLE DIFFERENCE DISCREPANCY MODELS ON IDENTIFICATION RATES

Adoption of regression over simple difference models for assessing severe IQ-achievement discrepancies may be hindered by uncertainty regarding differential model outcomes, particularly on rates and patterns of learning disability identification. Based upon previous research, this article identifies outcomes that could be anticipated following a change from a simple difference to a regression discrepancy model. Comparison of actual outcomes in determination of overall identification rates and the identification of students by race, grade, and time of evaluation (initial/reevaluation) are reported from a study of 194 referred students whose discrepancies were derived from both regression and simple difference models. The findings indicate that lack of bias across IQ levels raises identification rates for the regression model and makes identification more equitable across race and other variables when subgroups differ in mean IQ.