
Nate Cohn @Nate_Cohn
@davidshor @jon_m_rob @KevinReuning i think it's semi-persuasive, but it's not a great way around the aforementioned spatial autocorrelation problem. if we choose a 1000 random points in the Midwest, how often do we get an effect like that? county residuals map makes it seem like it could be fairly often! — PolitiTweet.org