Polytomous Cognitive Diagnosis Testlet Model 多级评分认知诊断题组模型

This research proposed the Polytomous Cognitive Diagnosis Testlet Model (PCDTM), which is suitable for dichotomous response data and polytomous response data and can also handle within-item multidimensional testlet effects. To investigate the performance of the PCDTM, we presented a simulation study and an empirical study. Results indicated that: (1)The PCDTM model is reasonable and effective, and accurate parameter estimation results can be obtained regardless of whether the answer data contain testlet effects or not; (2) as the sample size, item quality, and the number of item increase, the accuracy of estimation of the examinees' parameters, item parameters, and testlet parameters would be improved; (3) ignoring testlet effects will decrease the item parameter estimation accuracy and the examinees' accuracy rate, and even the estimation results may be confused. The results of the empirical study of the 2012 large-scale test in mathematics also show that the PCDTM model fits better and has accurate parameter estimation, which is more suitable for processing testlet data in actual data. This online resource is an R code of the PCDTM model with a specific Q, U, and Y matrix. Users can use this code to obtain the MCMC estimation of PCDTM to get the item, testlet effect, and examinee parameters.