Confidence intervals in repeated-measures designs: The number of observations principle.