A numerical majority of peer reviewed studies in the social sciences can't be replicated. Why is that?
