For peak energy, 0.3 of the error of measurement; for other measures, 0.2 of the between-subject standard deviation in Desk 2, adjusted for small-sample bias.