There is a raging debate right now about just how close the presidential race is. Is it really that close? Is it time for Democrats to panic? On and on and on. As you may have picked up from my tone I think there's a lot more heat here than light. Or to put it in statistical terms, the discussion has a high margin of ego and acrimony, even though the actual disagreements seem relatively limited.
The debate between Nate Silver and his critics essentially comes down to the meaning of "close". At heart there's an argument here over something that seems as emotive as statistical - how much should we emphasize that while Clinton remains the favorite to win, we can't take for granted or assume she'll win? This doesn't strike me as a truly statistical argument. More concretely, people are arguing over whether the current 538 model is more volatile than it should be, whether it's picking up a lot of polling noise as opposed to signal. This is a statistical and probability question that is frankly beyond me.
Read More →