Addressing score comparability in diagnostic classification models: an observed-score equating and linking approach

Abstract

The purpose of this study is to revamp and examine the observed-score approach for equating scores under diagnostic classification models (DCMs). The observed-score approach was adapted to accommodate the categorical latent traits and scores, a unique property of DCMs. Three simulation studies, each corresponding to a data collection design, were conducted to evaluate the amount of equating error and the improvement in score comparability with the observed-score approach. Findings indicate that (a) DCM scores are robust to form differences when high-quality items are used, and (b) the observed-score approach shows promise for yielding a small amount of equating error and increasing classification accuracy under small sample size conditions.