多维题组效应认知诊断模型

发布时间：2018-07-12 20:50

本文选题：认知诊断 + 题组　；参考：《心理学报》2015年05期

【摘要】：当前认知诊断领域还缺少对包含题组的测验进行诊断分析的研究,即已开发的认知诊断模型无法合理有效地处理含有题组效应的测验数据,且已开发的题组反应模型也不具有对被试知识结构或认知过程进行诊断的功能。针对该问题,本文尝试性地将多维题组效应向量参数引入线性Logistic模型中,同时开发了属性间具有补偿作用的和属性间具有非补偿作用的多维题组效应认知诊断模型。模拟研究结果显示新模型合理有效,与线性Logistic模型和DINA模型对比研究后表明:(1)作答数据含有题组效应时,忽略题组效应会导致项目参数的偏差估计并降低对目标属性的判准率;(2)新模型更具普适性,即便当作答数据不存在题组效应时,采用新模型进行测验分析亦能得到很好的项目参数估计结果且不影响对目标属性的判准率。整体来看,新模型既具有认知诊断功能又可有效处理题组效应。
[Abstract]:In the field of cognitive diagnosis, there is still a lack of research on the diagnostic analysis of tests containing problem groups, that is, the developed cognitive diagnostic models are unable to deal with test data with group effects reasonably and effectively. Moreover, the developed problem group response model does not have the function to diagnose the knowledge structure or cognitive process of the subjects. In order to solve this problem, this paper attempts to introduce multi-dimensional problem group effect vector parameters into the linear logistic model, and develops a cognitive diagnosis model of multi-dimensional problem group effect with compensatory effect between attributes and non-compensatory effect among attributes. The simulation results show that the new model is reasonable and effective. Compared with the linear logistic model and the Dina model, the results show that: (1) when the answer data contain group effect, Ignoring the item group effect will lead to the deviation estimation of item parameters and reduce the accuracy of the target attribute. (2) the new model is more general, even when the answer data does not exist the problem group effect. The new model can also get a good result of item parameter estimation and does not affect the accuracy of target attributes. Overall, the new model not only has cognitive diagnostic function but also can deal with problem group effect effectively.
【作者单位】：浙江师范大学心理系;北京师范大学认知神经科学与学习国家重点实验室;香港教育学院评估研究中心;
【分类号】：B842.1

【参考文献】