The direction of a comparison may be reversed when data for several groups are combined. For example, is it logically possible for every program in a university to accept a higher proportion of female than of male applicants, even though the university as a whole accepts a higher proportion of male than of female applicants?