Nate Silver @NateSilver538
It's actually not so small. There are hundreds of primary results since 1980, which is enough to empirically measure the shape of the probability distribution. Also, there are theoretical reasons to use fat-tailed distributions in cases of low n's. — PolitiTweet.org
Jon Walker @JonWalkerDC
The data set in terms of results is so small that is zero way to prove if this holds up or just provides more abili… https://t.co/0O6Zxzr70i