Abstract: Usually, cognitive diagnostic assessment (CDA) is based on a test and the corresponding cognitive diagnostic model to construct a diagnostic analysis. Many approaches need a Q-matrix which reflects how attributes are measured in each item when applying the cognitive diagnosis model into an assessment. Q-matrix plays an important role in CDA. Qmatrix can be defined by experts in related fields, and also can be estimated according to students' response data. Based on the existing Qmatrix refinement methods, a Q-matrix estimation method using an item fitting statistics RMSEA is proposed. The effectiveness and efficiency of the method are verified by a simulation study. And a real data analysis is also included. The results show that: (1) the CSE algorithm based on RMSEA can effectively estimate the attribute vectors of new items, and it takes less time; (2) the success recovery rate of Q-matrix estimation is greatly affected by the number of attributes and the number of basic items, especially when the number of attributes is large, it requires more basic items to estimate the attribute vectors of new items; (3) The sample size has little effect on the performance of CSE approach and a big sample size is not necessary to implement the Qmatrix modification method. Even if the number of subjects is 400, as long as the number of basic items is enough, it can have a high recovery ratio; (4) The application of this method to the analysis of empirical data can optimize the existing analysis results to a certain extent and improve the fitting of model-data.

Key words: cognitive diagnosis, Qmatrix estimation, item fit statistic, DINA model