I suspect that Mr. -Phipps meant to say “the jury is very much out,” thus implying that events may prove Mr. Silver’s forecast to have been in error (whereas “the verdict is very much out” can carry an implication that Mr. Silver has called the election for Obama, and will be proven wrong – particularly if one drops the “very”).
Mr. -Phipps, please correct me if I am mistaken.
Maybe, but that was also granted, Romney can still win, what I noticed in that post from **Phipps **was the affirmation that what silver is doing is mostly a “black box” when there is plenty of evidence to show what he has done.
IIUC what some are demanding from Silver is a little bit like demanding that he has to give the store away.
It is like affirming that there is no evidence that the makers of the famed Damascus steel never produced some of the best steel weapons of their day, when there is plenty of practical evidence that the things worked as the old warriors claimed and the sword makers of the day did show the general way they did them, but the sword makers of Damascus knew that they should not reveal all the secrets of their craft.
That logical limitation at hand, it is however clear that there is a lot one can check that Nate is doing properly, when NS investigated the odd data (and he realized it was odd thanks to his background) coming from Research 2000 others did check and confirmed the oddities and Research 2000 fell on its face because they could not explain properly when and how they did the specific data collection, pointing to fraud. In this case what Silver and others demanded was hardly a trade secret, so I do point out that affirming that NS methods are like a black box is leaving a lot of context out.
I may have used the wrong term - what I meant was that the jury is still out.
But I didn’t mean the jury verdict about Silver’s 2012 election call specifically. (As noted, I don’t think his model will been shown to be valueless simply by Romney winning, or validated by Obama winning). I meant the jury verdict as to the value of Silver’s model in general.
IMO the tract record of this model is too short, and too little is known of the nitty gritty details of the model, for it to be validated or rejected at this time. That would have to wait for a longer track record and/or a more in depth analysis of the methodology and assumptions.
Which was it? And how can we pretend to have a serious discussion about Nate Silver’s predictions when we can’t even agree on the most fundamental facts about them?
This. Despite NS’s snarky reply, is it true that RCP also was correct about 49/50 states in '08? If so, why should I believe him over RCP?
Agreed, another example is the way he adjusts polls after a convention. He has an “average convention bounce” assumption in his model. I can’t remember the exact figure, but it’s not important to my point. If you took the average convention bounce in the last 3 elections, it’d be different than if your average stretched back to the last 10 elections. Which one NS uses, and the reasons for his decision, are relevant and will affect the accuracy of his model, and they’re also largely a mystery, as you noted.
RCP is also predicting an Obama re-election.
I think I might have been wrong before about missing just one. According to this, he missed 3 Senate seats (not surprisingly, ones he had assigned a lower confidence level to).
The convention bounce is only temporary, and resolves itself after a week or two. He penalizes each candidate by a set amount after their conventions, a penalty which slowly goes to zero. It has no effect on the final projection results, i.e. his actual forecast before the election, which means it’s semi-useless in my opinion, but it does theoretically give an idea of what a candidate’s ‘true’ support is during the weeks immediately after their convention.
Also, a better way to measure his and RCP’s accuracy is by what % of votes each candidate was forecasted to receive, versus actual. I don’t have that readily available, but I believe that 538 was appreciably more accurate than any of his competitors.
This is far better, because 538 isn’t projecting who will win a given election, but rather forecasting election results for each race. If he forecasts ten candidates to have a 70% probability of winning, then you would expect three of them to actually lose. If none of them do, then he likely forecasted too high for these candidates, which reduces his accuracy. He isn’t ‘calling states’ like every other mainstream election predictor out there.
You shouldn’t necessarily and no one has indicated otherwise.
Are Silver and RCP in major disagreement? Minor disagreement?
Current RCP prediction no toss ups 290 Obama. 538’s mean prediction 299. So no not major electoral disagreement. Is that even minor?
Popular vote RCP calls “tie” right now while 538 calls it close, ranging day to day from 1.5 to 2% Obama.
I suspected as much but was too lazy to look it up. Thank you very much.
So I guess one can equally trust or distrust both of them.
He missed 2, not 3. That site has Alaska attributed incorrectly to the Dems - it took a while after the election for everything to be counted, but Lisa Murkowski (R) eventually won the seat. Here’s a recap of Silver’s results for that year, and you can compare it to the actual election here (note the red Alaska).
Your first cite says Nate Silver predicted a win by Joe Miller, the Tea Party backed Republican who won the Republican primary. But it was won by Lisa Murkowski, who was the incumbent was not on the ballot and ran as a write-in candidate. Nate was right about a Republican winning the seat, but chose the wrong Republican.
The worst posters -those who fabricate their data- never put out good polls. They might happen to be accurate by chance (though we would never know), but they never add information.
In contrast, systematically biased polls can add value provided you attach an adjustment factor. Polls with small sample sizes can also add value, though they should be weighted less. These are the sorts of things Nate does. He’s not working any miracles and someday I expect him to have many more competitors. For the present he is providing a valuable service for the reality based community.
Why the gay hate?
From astro? Because it’s a satirical reference to what Dean Chambers said (see the link) and not meant seriously.
From Dean Chambers? Because he’s an asshole.
Actually I take that back. Chambers didn’t mean to imply that Silver was gay when he said this:
He just meant that he’s subservient to women and liberals, and he also has no testicles.
A compelling rebuttal of Silver’s statistical analytical model, you must agree.
I agree he didn’t imply it. That’s some blunt, disgusting, old-fashioned gay bashing right there.
It’s sarcasm, implying that the objections conservatives have to Nate Silver has little to do with the validity or reliability or his model and rather to do with ideology.
I suppose “disproving” the effectiveness of Silver’s model is quite difficult. One of the components of experimental science is that following the same procedure one should receive similar (not different to a statistically significant degree) results. That may be why assigning a prediction interval to a one off event is so tricky - as rigorous as a model is, the conditions of any one election can not be replicated (or reproduced, as Drummond would have us say). Neither can one eliminate problematic variables. I hope that’s accurate anyway - I have had lectures on statistics, but I have learned quite a bit from this thread and would be glad for corrections.
Yep. my mistake - that race was still undecided the day after election day and I somehow decided to neglect it in the numbers (34/36, instead of 34/37), despite a decision having been reached two years ago. No good reason to do that, and I’m not sure why I did. Sorry!
Gang, the GOP routinely denies the existence of global warming, favors biblical creationism over scientific evolution, distrusts vaccines (especially when they have something to do with sex, such as HPV), and thinks God is creating hurricanes to punish us for allowing homosexuality.
They have a history of not allowing science to interfere with their antiquated world-view. In this context, I am not at all surprised the right is attacking Nate Silver; if this were 300 years ago, they would be putting him on trial for being a sorcerer and burning his demonic, future-predicting scrolls.