Investigating Anchor Set Purification in Test Equating

• In Event: Approaches to Scaling, Equating, and Linking

Abstract

This paper investigates anchor set purification processes in the context of 3PL IRT equating. It relies on a 3 x 2 x 2 unbalanced design with item parameter drift approach factor (null, d2 or robust z), screening rule factor (cut score value or critical value form standardized normal curve), and expert review factor (none or expert review). Factors effects are evaluated on item drift rates, scale transformation coefficients, and passing rates. Longitudinal data for multiple grades serve as population from which one hundred bootstrap samples were withdrawn for each grade and outcomes analyzed with descriptive and inferential statistics. Preliminary results show adequate performance of robust z. Final paper brings comparisons for all conditions and replications.

Author

• Alvaro J. Arce, Pearson