诸多研究显示用合成信度可以较好地估计测验信度。文献上对合成信度置信区间估计的研究都假设题目测量误差不相关,而在实证研究中,也会遇到误差相关的情况,此时α系数往往高估测验信度,使用合成信度估计测验信度比较准确。本文给出用Delta法计算一般的单维测验合成信度的标准误公式,此公式无论测验误差是否相关都适用,据此可以计算合成信度的置信区间。通过对600名青少年调查发现,中文版FAD分测验"总的功能"的反向题测验误差存在相关,演示了如何估计此分测验的合成信度及其置信区间。
Reliability is an important indicator in evaluating the quality of a test. Test errors are often correlated in some applied resear- ches, such as speed test, consistency response sets, transient errors and so on (Green & Yang,2009). In such case, based on confirmato- ry factor analysis, composite reliability can well estimate test reliability. We deduced a formula by using Delta method for computing the standard error of composite reliability of a unidimensional test in which the item errors could be correlated. Based on the standard error, the confidence interval of composite reliability can be obtained easily. We used the Chinese version of general function test of FAD as an example to illustrate how to calculate composite reliability and its confidence interval.