Multistage tests under D-scoring approach
Kyung T. Han, Dimiter M. Dimitrov, Faisal Al-Mashary · 2024
The D-scoring method for scoring and equating tests with binary items proposed by Dimitrov (2016 ) offers some of the advantages of item response theory (IRT), such as item-level difficulty information and score computation that reflect the item difficulties, while retaining the merits of classical test theory (CTT), such as the simplicity of number correct score computation and relaxed requirements for model sample sizes. Because of its unique combination of these merits, the D-scoring method has seen quick adoption in educational and psychological measurement. Because item-level difficulty information is available with the D-scoring method and item difficulties are reflected in test scores, it conceptually makes sense to use the D-scoring method with adaptive test designs such as multistage testing (MST). In this study, we developed and compared several versions of the MST mechanism using the D-scoring approach and proposed and implemented a new framework for conducting MST simulation under the D-scoring method. Our findings suggest that the score recovery performance under MST with D-scoring was promising because it retained score comparability across different MST paths. We found that MST using the D-scoring method can achieve improvements in measurement precision and efficiency over linear-based tests that use the D-scoring method.