Nate Silver @NateSilver538
@DanRosenheck @gelliottmorris @mattyglesias To get dorky for a moment: there are a *lot* of parameters to consider. The degree of the polynomial and the kernel function both matter quite a bit. So does whether, and how, you're adjusting for house effects. And then there are irregularities in the flow of data... — PolitiTweet.org