认知诊断模型资料拟合检验方法和统计量  

Model-data fit test methods and statistics for cognitive diagnostic models

作　　者：陈孚[1] 辛涛[2] 刘彦楼[3] 刘拓[4] 田伟[2] 

机构地区：[1]北京师范大学心理学院 [2]北京师范大学中国基础教育质量监测协同创新中心,北京100875 [3]曲阜师范大学中国教育大数据研究院,曲阜273165 [4]天津师范大学教育科学学院,天津300384

出　　处：《心理科学进展》2016年第12期1946-1960,共15页Advances in Psychological Science

基　　金：国家自然科学基金项目(31371047)资助

摘　　要：认知诊断模型界定了测验题目和所考察属性之间的关系,通过被试的作答反应获取被试对属性或知识技能的掌握情况。认知诊断模型资料拟合检验可以从项目拟合、模型绝对拟合、模型相对拟合和个人拟合方等方面进行。通过对认知诊断拟合检验方法和统计量的详细介绍和评价,可为认知诊断实践提供借鉴和参考。未来研究可在更丰富的研究条件下对各统计量的性能进行评价和对比,完善已有的拟合检验方法,提出新的拟合统计量。Cognitive diagnosis models （CDMs） are one family of psychometric models developed primarily for assessing examinees＇ mastery and non-mastery of skills or attributes. The extent to which the cognitive diagnosis models＇ hypotheses fit the real response patterns substantially influences the validity of cognitive diagnostic evaluation. The model-data fit tests in cognitive diagnostic theory are typically based on the item fit tests, the overall fit tests and the person fit tests. Comprehensive and structural introduction and evaluation of the model-data fit test methods and related statistics are intended to aid the future cognitive diagnostic studies and practice. Directions of future researches on model-data fit for cognitive diagnosis models include further inspection of the evaluation and comparison of fit test statistics under more complex simulation conditions, the improvement of the existing fit statistics and the development of novel fit statistics.

关 键 词：认知诊断模型 拟合检验 项目拟合 总体拟合 个人拟合 

分 类 号：B841[哲学宗教—基础心理学]