Yes. Nate Silver does large numbers of simulations. He reports predictions that converge.
I have no idea what that is supposed to do with the Bickers-Berry model under discussion. Silver made a bunch of off-the-cuff critiques without, I presume, having seen the paper since this month’s issue of the journal in question is still forthcoming.
Yep, that’s how we got the DNC and the Clintons.
I don’t know, anytime 100% of black people in the US are against something/ for something I’m not particularly optimistic about their preference. But Romney has been pretty darn incompetent.
I thought the result meant the sample was so small that it happened not to include any black Republicans (not an unlikely thing; a sample of 3000 tends to be enough). Hard to apply a statistical correction from 0 to any other number, right?
If you do a little digging you can find something that appears to be the paper in question on this page.
What do you mean, “not particularly optimistic about their preference”?
Five Thirty Eight does really good work. Nate Silver is one of the few political analysts out there who has the honest respect of both the left and the right. That’s impressive since he’s personally liberal. It’s very hard to keep bias out of your own work, and Nate Silver does it about as well as anyone else I can think of on the left or right.
My apologies. I was reading the discussion as being about the 538 model with the Bickers-Berry model being brought up as a compare and contrast example. I have to admit the digression to the discussion being specifically about the Bickers-Berry model still eludes my comprehension.
As to a side discussion about the University of Colorado Bickers-Berry model - I would completely agree that “validating” a model by showing that accounted for the data that you based the model off of is goofy. One can use post-dictions to test a model, but only if the data you are testing against was not used in model formulation - for example if they they only used the two last election cycles to develop their model and then went back and tested against previous cycles that were not used in model formulation and the data before the election also predicted the outcomes then that would be suggestive of the model’s strength. From what posters here are saying that is not what they have done.
In any case, my comment is the compare and contrast relevant to the op: 538 made predictions that did very well at the state by state and contest by contest levels. Given that what is being reported are probabilities my only complaint is that he was not “wrong” more … if he is stating that a particular contest has an 80% probability of turning out a particular way then, if his model is accurate, he should be wrong one out of five individual contest with those odds. Given that he was wrong in only two electoral districts and not a single Senate race his model is leaving too wide of standard deviation based on the data collected so far, unless he was making predictions with a 97% average probability (which he was not). If I hadn’t read his predictions ahead of time that data would be like the Mendel experiments … correct but so good that they had to be cooked.
Then the relevant trend line should be the trend for polls for that particular state - not national polls.
In some states there might not be enough state specific data for some states to provide a statistically significant trend line. For example he only cites 7 polls in Montana starting in November 2011. He refers to 32 polls in Ohio since December 2012 by comparison.
I guess the probabilities tell the frequencies output by his model, but don’t necessarily have to be interpreted as meaning “these would be the frequencies if you ran the election a large number of times.” Instead, they could be interpreting as meaning “Here’s how confident I am that person X will win this election.”
In other words, the assumption is that the frequency of wins within the model corresponds, not to frequencies of wins in the real world, but rather, to levels of confidence about predictions concerning the real world.
(On that reading, it would seem to me that the more Silver turns out right about individual elections, the closer to 100 and 0 the reported percentages should be in his future predictions–because the past record of success should rationally make him more and more confident about his results.)
Lamar Mundane is being sarcastic, moonshot925.
I just thought I’d mention that.
To you.
My wife just told me she heard that the Romney campaign was reducing its ad presence in Ohio and Pennsylvania. 1. Is this true? 2. If so, is it because they have reason to believe that these two states are shifting from “tossup” to “solid Obama”? 3. If that’s true, is part of their reasoning that the recent “defeat” of voter “suppression” in Ohio (via voting hours), and the possibility of true voter ID shenanigans in Pennsylvania being turned over on appeal, puts those states out of reach for the Repubs? (Or maybe, in the case of Pennsylvania, the opposite is true…they think the Voter ID crap won’t be overturned on appeal, and so they think Penn is now solidly Romney’s?)
You’ll never hear a campaign surrender before Election Night, but here’s the PA story.
Not the case in Ohio, though. Without PA, Romney *needs *OH, FL, and VA. All of them.
A correction and a question:
I phrased this poorly. Better than “prediction” might have been “fitting the training data” (or “post-diction” to use DSeid’s term). Note that some problems, perhaps including this one, come with such a meager amount of useful historical data that one may not have the luxury of setting some aside for testing.
Question:
Are Nate Silver’s predictions based on an election held today? The real danger for Obama is not noise in the polling results, or even normal voter shifting, but some unforeseen October Surprise that changes sentiment drastically. I think the possibility of a Surprise is why Intrade still gives Romney a 42% chance.
What will the Surprise be? I don’t know, that’s why they call it a Surprise! 
:smack: I meant DLC. Either yous knew what I meant or you didn’t catch my error.
But it won’t be Romney who springs it. He can’t; he has to wait for something to blow up in Obama’s face and somehow appear more competent than Obama by comparison.
Only an incumbent can spring an OS, and since it’s under his control, it means a bounce his way. But the argument is right if the reference is to mere chance events - and there’s only 8 weeks of window left for them to happen.
Thanks, Elvis. So you mean, the Romney campaign doesn’t think they have a shot in PA, with or without voter ID shenanigans…but as for Ohio, assuming the ad-spending pullback is true, what’s their strategy? Maybe saving up for a blitz late in the campaign?
They’re spending big in Ohio now, sez the article, and will keep spending right up to the end. Where have you seen anything about a pullback there? That would be suicidal.