摘要
在测量工作中,误用经典测验理论方法估计标准参照性测验的整体信度和等级线决策信度的情况非常突出。如,无论测量设计是交叉的还是嵌套的,也无论测验结果是做常模参照性解释,还是做标准参照性解释,测验工作者往往只报告克龙巴赫α系数或经典测验理论中的其它少数几个信度指标,而误把整体信度作为等级线信度的现象则更加普遍,这是十分不妥的。本文借用概化理论中的可靠性指数Ф和Ф(λ)公式,分别针对交叉设计和嵌套设计,就标准参照性测验的整体信度和等级分数线决策信度的估计问题进行了探讨。用数据演示的方法比较了交叉设计与嵌套设计在估计标准参照性测验整体信度方面的差异,展示了等级决策分数线决策信度的估计方法。
In criterion-referenced testing, dependability index based on generalizability theory is the correct indicator of measurement reliability. The present study presents the rationales and procedures for estimating dependability when the purpose of the test is to determine domain status of the individuals. Another important issue in criterion-referenced testing is how to determine reliability of different cut-scores. It is shown in this study that, reliability varies depending on the distance between a particular cut-score and the overall mean of the test. Cut-scores closer to the mean were found to have lower dependability than those farther away from the mean. It is also shown that the commonly used Cronbach ? is not an adequate index of reliability for criterion-referenced use of a test.
出处
《心理学探新》
CSSCI
2003年第3期52-56,共5页
Psychological Exploration