Invariance of Score Linkings Across Gender Groups for Forms of a Testlet-Based College-Level Examination Program Examination

Abstract

This study investigates whether the functions linking number-correct scores to the College-Level Examination Program (CLEP) scaled scores remain invariant over gender groups, using test data on the 16 testlet-based forms of the CLEP College Algebra exam. To be consistent with the operational practice, linking of various test forms to a common reference form is based on the Rasch model. Equatability indices proposed by Dorans and Holland (2000) were used to evaluate linking invariance over gender subpopulations. Overall, linkings based on the gender groups are very similar to linkings for the total group. At all score levels, differences between subgroup and total group linkings are smaller than the difference that will affect the pass/fail decision for CLEP candidates. On only one form, linking based on the male group would pass slightly more candidates than linking for the total group at the recommended CLEP cut score of 50 due to rounding.