gwern comments on Why epidemiology will not correct itself