A Multi-strategy Cognitive Diagnosis Model Incorporating Item Response Times

Cognitive diagnosis aims to deeply explore the internal mental activity rules and processing mechanism of the individual,and diagnose the individual from a microscopic perspective,so as to accurately measure the individual cognitive process and provide corresponding remedial methods.In the process of cognitive diagnostic evaluation,cognitive diagnostic model plays a key role,and the quality of the model determines the accuracy of cognitive diagnostic evaluation.Most cognitive diagnostic models can only deal with single-strategy test situations,but individuals often use multiple strategies when solving cognitive tasks,which leads to incorrect model Settings or inadequate fitting,thus affecting the validity of inference.Existing strategy more cognitive diagnosis model in modeling only consider the individual response accuracy,while strategy choice theory believes that individuals will weigh the speed and accuracy of strategies to choose the optimal strategy and response time is very useful in analyzing the strategies used by individuals when solving problems.Based on this,this paper,based on the development idea of GMS-DINA model(Ma & Guo,2019),considers the response time in the modeling of strategy choice.In other words,a multi-strategy cognitive diagnosis model combining response time information is developed(RTMS-DINA),and the corresponding parameter estimation is carried out by using MCMC algorithm.In order to test the performance of the new model,this paper carries out theoretical research,simulation research and empirical research respectively.Theoretical research is to explore the development process of new model from the theoretical perspective;Two simulation studies are used to illustrate the effectiveness and robustness of the RTMS-DINA model.The empirical study is used to explore the fit of RTMS-DINA model in Raven Advanced Reasoning Test data.The results show that:(1)RTMS-DINA model has high accuracy in parameter estimation and classification accuracy of attributes.Compared with the uniform distribution,the accuracy of parameter estimation and attribute classification was better when the attribute profile was taken from the high-order ability distribution.(2)The GMS-DINA model and MS-DINA model fit the data generated by the RTMS-DINA model,and it was found that the RTMS-DINA model always obtained the most accurate attribute classification,while the MS-DINA model had the worst attribute classification accuracy.(3)Empirical studies showed that RTMS-DINA and GMS-DINA had the same degree of fitting to Raven Advanced Reasoning Test data;The reliability of RTMS-DINA model in attributes was better than that of GMS-DINA model;Compared with the inferences of the GMS-DINA model,the inferences of the RTMS-DINA model are closer to the conclusions of the observation of the response.Therefore,the multi-strategy cognitive diagnosis model RTMS-DINA is more suitable for practical application by combining response time information.