Effect of regression to the mean in the presence of within-subject variability.