Statistics notes: Treatment allocation in controlled trials: why randomise?