A Polytomous Cognitive Diagnosis Model Based On Partial Credit Theory

Traditional educational tests give the test scores,the purpose is only to rank selection or assessment of the overall level,regardless of the potential psychological process,cognitive characteristics.Unlike the traditional test,the Cognitive Diagnosis Assessment(CDA)can further provide information on the individual knowledge structure or processing skills,and the educator can follow the information to teach remedies to meet the individual needs of students.The implementation of CDA need some excellent performance and can promote the use of cognitive diagnostic model(Cognitive Diagnosis Model,CDM),so the development of the model is particularly important.CDA development process,the emergence of a large number of models for different test situation,but most models are limited to measuring 0-1 scoring questions,the problem of multi-level scoring can only be converted to 0-1 score way to deal with,This way to lose a lot of useful information.For example,for a multi-level score,the subject needs to complete multiple steps of the answer,in each step the subject may need to call different cognitive attributes to complete the problem,if the use of multilevel scoring CDM,you can accurate estimation of the subject knowledge structure.In the development of educational test,the idea of extending 0-1 scoring into multi-level scoring has been developed in the development of Item Response Theory(IRT),which is typical of Graded Response Model,GRM)and Partial Credit Model(PCDM).The existing multi-level score cognitive diagnosis model,mostly based on the idea of hierarchical response model developed.In this study,the cognitive diagnosis of DINA model is extended to multi-level model(hereinafter referred to as PC-DINA),and the model and the domestic development of GRM according to the P-DINA model,sequential processing of multi-level The diagnostic model(Sequential DINA model,SEQ-DINA)was compared.Research indicates:(1)The newly developed PC-DINA model has high accuracy,good model robustness and good fitting of model data,which indicates that the model is reasonable and feasible.(2)The comparison between PC-DINA,P-DINA and SEQ-DINA shows that the accuracy of parameter estimation is influenced by the quality of the true model in both the coincidence and inconsistency of the true model and the fitting model.The accuracy of parameter estimation is also reduced,and the accuracy of parameter estimation is also affected by the number of subjects.When the number of subjects is reduced,the accuracy of parameter estimation will be reduced,but the decrease is small.When the parameter quality is low,the judgment rate is obviously decreased,and the judgment rate is not obvious.(3)PC-DINA,P-DINA,SEQ-DINA mutual verification study shows that the model is consistent with the fitting model,the model data fitting index is the best,when the true model and the fitting model is not consistent,the model fitting index Will be down.(4)When the fitting model agrees with the true model,the fitting of PC-DINA is the best,followed by the P-DINA model,and the SEQ-DINA model is the worst.When the model is not consistent,the SEQ-DINA The overall performance of the model is the best,followed by the P-DINA and PC-DINA models.(5)The use of real data shows that the PC-DINA model is best fitted with the PDINA model and the SEQ-DINA model is the worst.