Vonesh E F, Schork M A
Biometrics. 1986 Sep;42(3):601-10.
Determination of sample sizes for comparing two or more treatments in repeated measurements experiments is considered. Multivariate normality of the individual's vector of repeated measures is assumed. Particular emphasis is placed on applications wherein the error variance-covariance matrix is arbitrary positive-definite. Sample size determination is based on power considerations associated with Hotelling's T2 test and the desire to detect a specified difference between any pair of treatment means. Tabulated sample sizes are given in the case of an equal variance-unequal covariance structure. The utility of these sample sizes is also demonstrated for a more general variance-covariance structure. Applications of these sample sizes are illustrated with two examples.
考虑在重复测量实验中比较两种或更多种处理的样本量确定问题。假设个体重复测量向量服从多元正态分布。特别强调误差方差 - 协方差矩阵为任意正定矩阵的应用。样本量的确定基于与霍特林T2检验相关的功效考量以及检测任意一对处理均值之间特定差异的需求。在等方差 - 不等协方差结构的情况下给出了列表形式的样本量。对于更一般的方差 - 协方差结构，也证明了这些样本量的实用性。通过两个例子说明了这些样本量的应用。
