Development Of A Saturated Cognitive Diagnosis Model That Incorporates Response Time

Cognitive Diagnosis(CD)aims to deeply explore the law of psychological activities and processing mechanism within individuals,so as to achieve accurate measurement of individual cognitive processing and provide targeted remedial measures.Cognitive Diagnosis Models(CDMs)as the key of cognitive diagnostic evaluation,are mainly used to reflect the relationship between individual's attributes and the response of items.The quality of CDMs determines the accuracy of cognitive diagnosis evaluation.Due to the complexity of the actual test,researchers need to develop a variety of models to meet the practical needs,which makes the development of cognitive diagnosis model become a hot topic at home and abroad.However,most of the currently cognitive diagnostic models only consider the individual's responses,and another important information-Response Times(RTs)is often ignored by researchers.With the development of computer,the item response time data has been easily accessible.Based on this,inspired by the hierarchical modeling framework of van der Linden(2007),this study intends to develop a saturated cognitive diagnosis model(JVRT-LCDM)that incorporates response time and apply MCMC(Markov Chain Monte Carlo)parameter estimation algorithm to estimate model parameters.In order to verify the performance of the new model,three studies have been carried out,namely a theoretical study and two experimental studies.Specifically,the theoretical study mainly discusses the development process and the corresponding parameter estimation process of the new model.Two experimental studies are divided into empirical study and simulation study.The empirical study is utilized to explore the model fit of JVRT-LCDM in real data,and to examine the relationship between the variable speed of subjects and their ability.The simulation study is used to verify the rationality of the estimation algorithm,investigate the effectiveness of JVRT-LCDM and explore whether the number of subjects will influence the estimation accuracy of JVRT-LCDM.The results show that:(1)Compared with its reduced model,JVRT-LCDM fits better in real data,which indicates that JVRT-LCDM has good applicability in practical application.(2)The speed of the subject is variable during the whole test.There is a moderate negative correlation between subject's ability and initial speed,a positive correlation between subject's ability and linear trend and a negative correlation between subject'sability and quadratic term.Moreover,the speed trajectories of high ability subjects are significantly different from those of low ability subjects.(3)The estimation of JVRT-LCDM is accurate.The recovery of parameters is greatly affected by the number of subjects.The larger the number of subjects is,the higher the accuracy of the parameter estimation is.