ABSTRACT
A review of the statistical issues surrounding power is warranted to set the stage for the discussion of estimating sample size and associated power for repeated-measures analysis of variance (ANOVA). Moher et al recommends that all studies report sample-size calculations, including primary outcome variable; clinically important treatment effect; absolute or relative status of this effect, and statistical test, directionality, alpha level, and statistical power used to estimate sample size. To determine appropriate sample size for a repeated-measures ANOVA, two approaches are used. The first is to perform a univariate sample-size calculation based on the outcome expected at a single point in time. The second is to 'average' across the repeated measurements by dividing the variance anticipated at any time point by the number of repeated measurements. Therefore, it is important from fiscal and ethical perspectives to match the design and hypothesis to the sample-size estimation technique.
