Answers are grounded in this paper and cite their sources.
Compares estimation methods for testlet-based cognitive-diagnostic models, supporting Bayesian approaches.
Testlet-based cognitive-diagnostic models involve both discrete attribute profiles and continuous testlet effects, making person-parameter estimation more challenging than in standard cognitive-diagnostic models. This study compares maximum likelihood estimation (MLE), maximum a posteriori (MAP), and expected a posteriori (EAP) under the IT-DINA model. A simulation study was conducted to evaluate the recovery of testlet effects and the accuracy of attribute classification under varying testlet lengths, numbers of testlets, and parameter settings. Furthermore, an empirical analysis using TIMSS 2023 mathematics data was carried out. Results showed that, under the studied model specifications and simulation conditions, Bayesian methods provided more stable recovery of continuous testlet effects than MLE. For attribute-pattern classification, Bayesian methods achieved higher accuracy in the baseline and moderately misspecified conditions, whereas their advantage weakened when the assumed attribute-profile prior was inconsistent with the true population distribution. In the empirical analysis, the IT-DINA model showed better relative fit than the DINA model and produced different person-level diagnostic classifications. These findings support considering Bayesian estimation in testlet-based cognitive diagnosis and highlight the importance of accounting for testlet effects in practice.
No takes yet. Share an insight, caveat, or question.
You et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: