影响AHM与DINA诊断准确率的因素研究

The Research on Factors Influencing Diagnostic Accuracy in AHM and DINA

对属性层级模型(AHM)和确定性输入、噪声"与"门模型(DINA)2个认知诊断模型,讨论不同因素对判准率的影响.实验表明,含有可达阵的测验比不含可达阵的测验判准率高.对于线型结构测验长度达到一定值以后,增加测验长度对诊断准确率的改进不大.虽然总体而言,DINA分类准确性要优于AHM,属性结构紧密度越大,判准率越高;但是AHM估计结果符合属性层级结构,而DINA估计结果却可能违背属性层级关系,这和2011年De Carlo的发现不相同.

Cognitive diagnostic theory is the product of combination cognitive psychology with modern psychological and educational measurement and it is a core of the new theoretical generation. Various factors, such as attribute hierarchy, test blueprint and test length, and the quality of the items, i.e., the quantity of the slip attached to an item will affect the accuracy of diagnosis to some extent. Based on two diagnosis models AHM and DINA, affecting degree of different factors to accuracy index of classification is investigated in the paper. The results show different factors have varying degree of influence on the accuracy of diagnosis. Test-length is not the longer the better when attribute structure is linear. Different test blueprint has varying degree of influence on the diagnostic accuracy. A cognitive test that contains reachability matrix has a higher classification accuracy rate than that of not contained. The higher the slip, the lower the pattern and the marginal classification accuracy rate are. The more compact the attribute hierarchy, the higher classification accuracy rate. AHM is more sensitive to attribute construction, while DINA is not. Generally speaking, the accuracy of classification of DINA is higher than AHM, but sometimes DINA model may provide some unreasonable estimate of knowledge states. This finding does not coincide with DeCarlo＇s conclusion.