Articulation of Cut Scores in the Context of the Next-Generation Assessments

Abstract

Logistic regression (LR)-based methods have become increasingly popular for predicting and articulating cut scores. However, the precision of predictive relationships is largely dependent on the underlying correlations between the predictor and the criterion. In two simulation studies, we evaluated the impact of varying the underlying grade-level correlations on the resultant bias in cut scores articulated using the LR method. In Study 1, we compared different articulation methods (LR and equipercentile smoothing), and in Study 2, we evaluated different criteria for linking (e.g., adjacent grade or end of course). The collective results indicate that as correlations became smaller, cut scores articulated using LR-based predictions became increasingly biased when compared to a true value obtained under perfect correlation. The predicted values are significantly biased for lower achievement levels, irrespective of the linking criteria used. Results from these studies suggest that the LR method must be used with caution, particularly when articulating cut scores for multiple achievement levels.