In defence of average cohorts