“There are so many ways to account for negative outcomes that it is safer to doubt one’s methods before doubting one’s subjects.”