层级相合性指标的修正和扩展.doc

  层级相合性指标的修正和扩展# 丁树良1，毛萌萌2* （1. 江西师范大学计算机信息工程学院，南昌 330022； 5 10 15 20 25 30 35 40 45 2. 南昌大学教育学院，南昌 330031） 摘要：Cui 和 Leighton（2009）提出个人拟合指数（HCI）以探测被试观察反应模式和期望 反应模式的拟合程度。但是 HCI 的计算公式中的分母可能为零；HCI 并没有考察被试在子 项目错误作答，而在父项目上正确作答的失拟的情形。针对以上缺点，本文对原始 HCI 提 出了修改和拓展并提出了新的 HCI （NHCI）。模拟实验结果显示 HCI 和 NHCI 在不同的 情况下各有优势。本文还认为 HCI 可应用于所有非补偿型的认知诊断模型（CDM），即对 CDM 不太敏感而与 Q 矩阵关系密切。不在 Q 矩阵的基础上考察 HCI 是无意义的。因此， 在施测前探测测验 Q 阵同认知结构是否一致是很重要的。文中提出了用于描述认知诊断测 验理论结构效度指标。 关键词：认知诊断；认知模型；认知诊断模型；层级一致性指标；理论有效性 中图分类号：B841 Modification and extension of the hierarchy consistency index Ding Shuliang1, Mao Mengmeng2 (1. School of Computer Information Engineering, Jiangxi Normal University, NanChang 330022; 2. School of Education, Nanchang University, NanChang 330031) Abstract: Cui and Leighton (2009) proposed the hierarchy consistency index (HCI) to detect the fitness of an examinee’s observed response pattern (ORP) to an expected response pattern (ERP) as a person fit index.HCI is not defined well in the sense that when an examinee mastered an attribute only and there is an item with the same attribute in the test, and the examinee responses correctly to the item only. The original HCI could not be computed for the number of comparisons being zero, hence the denominator being zero.Unlike the person fit indices in the item response theory, HCI is not heavily dependent on the cognitive diagnostic model (CDM) provided that the CDM is the non-compensable, but the HCI is heavily dependent on the test specification, the Q matrix. The HCI is meaningless unless the Q matrix is detected. In the paper there is an example (Sinharay and Almond, 2007) to demonstrate the conclusion. The example shows that 24 knowledge states are obtained from the theoretic cognitive model but the test specification could distinguish only 9 knowledge states. The test specification does not really coincide with the cognitive model. It is very important to detect whether a test specification coincides with the cognitive model before adm