标定Q矩阵是认知诊断评估中最基本也是最为关键的一步.如今Q矩阵标定的统计方法,多数为验证性方法,即验证或修正已有Q矩阵中元素的方法.在常见的Q矩阵未知和已有作答数据情形下,提出将探索性因素分析方法和验证性方法相结合的Q矩阵标定方法,并采用模型整体拟合指标、分类准确性指标等,综合确定属性数和Q矩阵.模拟研究表明:新方法可较好标定Q矩阵.
The vast majority of the existing statistical methods to specify the Q-matrix do rely on the draft Q-matrix constructed by the subject matter experts or the researches. Based on previous related research,the authors consider the estimation problem of the Q-matrix under the condition of without having the draft Q-matrix. In particular,they introduce a Q-matrix identification approach which integrates exploratory factor analysis( EFA) with Q-matrix validation method. The results of a simulation study show that the initial Q-matrix can be correctly explored with high recovery rate using the EFA and the validation methods.