Sample Size Calculations for Two-Group Repeated-Measures Experiments

Vonesh and Schork (1986, Biometrics 42, 601-610) presented a statistical methodology for computing the minimum sample size required for the within-subjects repeated-measures design. They applied Hotelling's T2 analysis and demonstrated the utility of these techniques under general covariance structures. In this paper, we extend these procedures to the between-subjects repeated-measures design when there are two treatment groups under consideration. The multivariate analysis of variance approach to analyzing repeated measurements is considered and this model also resolves to Hotelling's T2 analysis. Two models are put forward to contend with those situations where the correlation structure among the repeated measures is unknown. These include the autoregressive model and the assumption of compound symmetry, and they are compared and contrasted throughout. Tables of the minimum sample sizes required for several hypotheses arising from repeated-measures designs are presented.