The merits of matching in community intervention trials: a cautionary tale.