英语听力认知诊断测评模型优化研究  

Optimizing the Q-Matrix for a Cognitive Diagnostic EFL Listening Test

作　　者：董艳云[1] 马晓梅[2] 孟亚茹[1] DONG Yanyun;MA Xiaomei;MENG Yaru(School of Foreign Studies,Xi'an Jiaotong University,Xi'an 710049,P.R.China)

机构地区：[1]西安交通大学,陕西省西安市710049 [2]西安交通大学外国语学院,陕西省西安市710049

出　　处：《现代外语》2020年第3期389-401,共13页Modern Foreign Languages

基　　金：国家社科基金一般项目“英语认知诊断测评模式构建及有效性论证”(17BYY015)的阶段性成果。

摘　　要：本文旨在探究如何根据认知诊断模型拟合方法优化英语听力诊断性Q矩阵。首先对已有Q矩阵属性标注进行可靠性验证及人员拟合性分析,发现原Q矩阵仍有优化空间,其次采用DRFS法:拆分解析-匹配组合以及基于G-DINA测量模型的量化拟合筛选,进一步在原Q矩阵模型基础上优化出与数据拟合更佳的模型。结果表明:第一,优化的新模型在数据的相对拟合值和对分数变异的解释力及诊断力上好于原有模型;第二,新模型生成的属性掌握情况与被试自评结果的相关性更高,表明其属性掌握概率更接近自评结果。本研究提出的DRFS优化方法有望弥补以往Q矩阵构建及筛选的不足,为准确的模型构建提供借鉴。This paper attempts to optimize the EFL listening Q-matrix based on cognitive diagnostic model fit method.The expert judgment agreement and person-fit analysis showed that the original Q-matrix needed to be improved.Then the Disassemble-Reconfigure-Fit-Screen(DRFS)method was proposed:the 7 experts Q-matrices were disassembled and re-configured based on different internal consistency coefficients of items,and then their G-DINA model-data fits were compared to screen out the poorly fit models and obtain the best fit one.The results show that:1)the optimized model’s relative fit is better than that of the original model and the new one has more advantages in variance explanation and ability diagnosis;2)the subjects self-assessments are more consistent with the optimized attribute mastery patterns.This DRFS optimization method is intended to address the limitations of qualitative approaches in Q-matrix validation and provide reference for model construction.

关 键 词：认知诊断方法 英语听力 一致性检验 Q矩阵模型优化 

分 类 号：H319[语言文字—英语]