Pretty much every conclusion drawn from all studies where P < .05 is used as a proxy for truth, but which are later incapable of replication.

https://news.ycombinator.com/item?id=20895433