Comparative Research On The Diagnostic Results Of Latent Trait Model And Latent Class Model

According to the different types of skill mastery scale, Cognitive Diagnosis Models could be divided into two categories: latent trait models and latent class models. The former represents an attribute with continuous variables, while the latter represents an attribute with discrete variables, marked with 1 and 0. In practical applications, only one cognitive model is used to analyze the candidates' test scores generally. Using latent class model, the user can only get the information that whether the candidate masters the attribute or not, but without more detailed information, such as mastery degree remain unclear. Although the user can know the candidate's ability of each dimension by latent trait model, he could not accurately decide whether the candidate has mastered the attributes. Hence, relying on only one diagnosis result, the actual users of Cognitive Diagnosis will not get enough information. It will not provide effective guidance or help for teachers and students. Therefore, this study is to explore further information from the perspective of comparing the Diagnostic Result of Latent Trait Model with the Diagnostic Result of Latent Class Model.In the Monte Carlo simulation study, two experiments were set up to simulate a certain number of candidates to take an exam, which consists of a number of items and attributes. Attributes Hierarchy is independent. This paper was aimed to explore the correspondence of diagnosis results between the latent trait model(CCM, MIRT-NC) and latent class model(DINA), with analysis on the scores of fractional operation of Eighth-grade mathematics by Nie Bin(2009). And the main conclusions are as follows:(1) Different test answer data can affect the result of the comparison of the Diagnostic Result of Latent Trait Model with the Diagnostic Result of Latent Class Model.(2) For the three simulated test answer data by different models, with the results of CCM, DINA model analysis, it can be found that the proportion of attribute mastery of DINA model, increased in accordance with the order of(0, 0.2],(0.2, 0.4],(0.4, 0.6],(0.6, 0.8 ] and(0.8, 1.0].(3) For the three simulated test answer data by different models, with the results of CCM, DINA model analysis, it can be found that in(0.6, 0.8] interval and(0.8, 1.0] interval, attributes mastery property is greater than the proportion of attributes nonmastery; in(0, 0.2],(0.2, 0.4],(0.4, 0.6], the proportion of attributes mastery is less than the proportion of the attributes which is non-mastery.(4) In the simulation and actual data analysis, by CCM, DINA model comparison, it can be found the proportion of attribute mastery of DINA model is largest within(0.8, 1.0] interval, and obviously higher than other intervals.(5) In the simulation and actual data analysis, by MIRT-NC, DINA model comparison, it can be found the proportion of attribute non-mastery of DINA model is largest within(-0.6, 0.6] interval, and obviously higher than other intervals.(6) In the simulation, the ability value within(-1.8, 1.8] interval contains all cases of attribute mastery and attribute non-mastery by DINA model; in actual data analysis, the ability value within(-0.6, 0.6] interval contains all cases of mastery attribute and non-mastery attribute by DINA model.