Hacker News Reader

From what I've heard, Nate Silver received publicity for picking 49 out of 50 states in the 2008 election. I'd like to see how his model fares in 2012 and beyond, or a higher confidence interval at the very least. Is it falsifiable? Then we might be talking about science.

View on HN

What the hell do you mean "is it falsifiable"? He's making concrete predictions about the outcome of an event that is mere days away. Voters will have the opportunity to select an outcome different from what's been predicted, so it's possible for him to be wrong. Is there some other definition of falsifiable that you are using?

It's even better than that: by looking at his state-by-state predictions and comparing them to election results, we'll be able to quantify his degree of wrongness.

You're not implying that a Romney win would falsify his model, are you?

I don't think he has a specific "model" for his "predictions". Reading his recent article, he uses statistics and weights informations gathered by polls to determine the probability of each candidate winning per state. You can't really be "wrong" in that sense, but the actual result can vary due to statistical sampling error, polling error, or bias in the polls.

Source: fivethirtyeight.blogs.nytimes.com/2012/11/03/nov-2-for-romney-to-win-state-polls-must-be-statistically-biased

Reply on news.ycombinator.com