THE BIASING EFFECTS OF VIOLATING THE INDEPENDENCE ASSUMPTION UPON THE POWER OF t TEST
THE BIASING EFFECTS OF VIOLATING THE INDEPENDENCE ASSUMPTION UPON THE POWER OF t TEST
复制标题
违反独立性假设对 t 检验功效的偏差影响
DOI：
10.5926/jjep1953.44.2_234
复制
发表时间：
1996
影响因子：
--
通讯作者：
Kayoko Kurita
中科院分区：
文献类型：
--
作者：
Kayoko Kurita
在教育和心理学研究的许多领域，t检验是最流行的统计分析方法之一。 t 检验的假设之一是观察的独立性，这很重要，但在实际数据中很少令人满意。 t 检验的功效分析还需要独立性假设。违反此假设的影响已针对 I 类错误的概率进行了广泛研究，但并未针对功效进行研究。本研究的目的是通过计算机模拟的方式研究观察的非独立性对 t 检验功效的偏差影响。在本研究中，观测值的非独立性是由观测值子组内的类内相关系数定义的。结果表明，偏差的大小可能相当大，具体取决于总体效应大小、组内相关系数和子组的大小。
In many areas of educational and psychological research, the t test is among the most popular statistical analysis methods. One of the assumptions for the t test is the independence of observations, which is important but rarely satisfying in real data. The power analysis for the t test also requires the independence assumption. The effects of the violation of this assumption has been extensively studied with respect to the probability of Type I error, but not regarding the power. The purpose of this study is to investigate the biasing effects of non-independence of observations on the power of the t test by means of computer simulation. In this study, the non-independence of observations is defined by the intraclass correlation coefficients within subgroups of observations. The result suggests that the magnitude of the bias can be considerable depending upon the population effect size, intraclass correlation coefficient, and the size of subgroups.