It is a capital mistake to make empirical generalizations before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts.