Non-Parametric CD-CAT Item Selection Strategy and Termination Rules Based on Binary Search Algorithm

Abstract

CD-CAT plays a significant role in diagnosing and assessing students, contributing significantly to improving teaching effectiveness. However, in classroom teaching scenarios, unlike large-scale assessments where a large number of samples can be used to accurately estimate item parameters, non-parametric CD-CAT becomes the only feasible choice. Compared to parametric CD-CAT, non-parametric CD-CAT started later, and research mainly focuses on non-parametric item selection strategies. However, the existing non-parametric item selection strategies have the disadvantage of low efficiency, and there is little research on non-parametric termination rules. Therefore, this study proposes two more efficient item selection strategies: Non-Parametric Dynamic Binary Search (NDBS) and General Non-Parametric Dynamic Binary Search (GNDBS), as well as a non-parametric termination rule：Non-parametric Dynamic Binary Searching Index (NDBI). Simulation results show: (1) Under all conditions, the pattern classification accuracy rate of NDBS is higher than that of NPS, so NDBS can be used as the item selection strategy when there are no samples available. (2) In most cases, the performance of GNDBS is better than other item selection strategies, so GNDBS can be chosen as the item selection strategy when there are few samples available. (3) In variable-length tests, when the research objective is to obtain more accurate classification results, the critical value of the NDBI rule can be reduced; conversely, the critical values of the NDBI and GNDBI rules can be appropriately increased.