可以处理补偿作用的认知诊断模型:CDINA模型

对DINA和DINO模型进行改进,构建了一个比DINA和DINO模型更"一般"的模型,称为CDINA模型.除了猜测和失误参数之外,在CDINA模型中每个项目还包含1个补偿参数.构建的模型不仅保留了DINA模型简单和易于解释的特点,而且模拟和实证研究结果表明:(i)CDINA模型可同时处理项目属性之间是连接、部分补偿或完全补偿的情况(可将DINA和DINO看成是CDINA的特例),且CDINA模型具有较高的参数估计精度和被试分类准确率;(ii)通过比较CDINA、DINA、DINO模型的分类判准率和相对拟合指标,发现CDINA不会比DINA和DINO的分类准确率更低,且在存在部分补偿的情况下CDINA模型更优于DINA和DINO模型.

The Cognitive Diagnostic Model Which Can Deal with Compensation and Noncompensation Effects:CDINA

The DINA and DINO models are improved to build a more “general” model that is called CDINA model.In addition to the guessing and slipping parameters,each item in the CDINA model contains a compensation parameter.The model built preserves the parsimonious and easy-to-explain characteristics of the DINA model,and the simulation and empirical results show that the CDINA model can handle the item with its attributes are conjunctive,partially compensated,or fully compensated.That is,CDINA can be regarded as a general case of DINA and DINO,and it has relatively high parameter estimation accuracy and classification accuracy.The classification accuracy of the CDINA model is evaluated by comparing the three models of interest based on the relative fit indicators,CDINA is found to be no less accurate than DINA and DINO classifications.In the case of partial compensation,CDINA outperforms the DINA and DINO models.