Adaptive Testing for Multiple-Choice, Multiple-Attempt Test Items

This study explores the integration of multiple-choice, multiple-attempt test items within the Computerized Adaptive Testing (CAT) framework, termed as MM-CAT. A simulation study was conducted to investigate the effects of integrating multiple attempts on items in the CAT environment. The findings from this study are twofold. First, we observed that the allowance of multiple attempts on test items significantly enhances the accuracy of $\theta$ estimates within fixed-length CAT settings. Second, our results indicate a noticeable reduction in the necessary test length for variable-length CAT when multiple attempts are permitted. From these results, given the same item bank size, MM-CAT outperforms the traditional CAT which uses the correct/incorrect scoring.