The most pressing question isn’t whether Silver is an effeminate gay idiot or whether liberals rely too heavily on his opinion, but rather what 538’s logo is meant to denote. I’ve held that it’s a very patriotic duck. My friends have recently tried pointing out that it’s actually a calculator, which is shocking if true.
I used to see a calculator, but now that I’ve seen the nightmarish three-eyed ultra-American duck, I can’t unsee it.
I strongly disagree. If they published their methodology, what more do you want? Their hypotheses will be proven false but as long as they own up, look at their system honestly, and fix some things, the actual analysts (neither Colorado nor Nate Silver) shouldn’t be discredited.
The Colorado prediction lives in a world where people know that, but even Nate Silver has built up enough of a reputation such that, if he’s very wrong but still desires to, he’ll have another respected prediction operation again in 2016.
Andrew Gelman has a nice piece in the NY Times about what too close to call really means:
This is pretty much required reading if you’re at all interested in probabilistic forecasting.
This may show the sad state of my psyche, but I always see it as a bed. A sexy, sexy statistics bed. Bow chicka bow.
Of the 37 Senate seats contested in the November 2, 2010 elections, 36 were resolved by November 4, including very close outcomes in several states. Of these 36, the FiveThirtyEight model had correctly predicted the winner in 34.
Of the 37 gubernatorial races, FiveThirtyEight correctly predicted the winner of 36.
In final vote tallys as of December 10, 2010, the Republicans had a net gain of 63 seats in the House, 8 more than the total predicted on election eve though still within the reported confidence interval.
Cite: FiveThirtyEight - Wikipedia
So if we’re charitable and say he missed the House race by 8, that’s a total of ten errors in 508 tracked races. Or an accuracy rate of 98%.
That’s your definition of “just okay”? You’re worse than the Asian Dad meme.
Dammit! You got me, too!
You’re mixing and matching numbers in an utterly screwy way here.
He was 34 / 37 in the Senate, right? that’s 91.9%
In the governorships, he was 36 / 37, so that’s 97.3%
In the House, predicting how many seats the Republicans will pick up is not remotely the same as predicting the winners of individual House races. You can’t just combine the Senate and Governor’s mansions (where he did pick individual winners of particular races) with his House number and call them all “tracked races”. Not even close. It’s not that he correctly forecasted the winner in 427 out of 435 races. He took his best estimate of how many House seats the Republicans would pick up, knowing full well that some that he thought were likely pickups would not happen, and some that he thought were leaning Dem would fall.
You can’t lump the Houses 435 races in with “tracked races”, and you certainly can’t give him credit for calling 427/435 of them correctly when, to my knowledge, he didn’t predict a winner in any individual House races.
Fair enough. I couldn’t find his 2010 Rep predictions but I wasn’t sure if it was because he didn’t have any or because my Google-Fu wasn’t strong enough. Turns out it’s the former.
He’s still at 94.5% accuracy in his selections for 2010. Better than “just okay.” Better than Rasmussen (34/37 in Gov, 33/37 Senate, 90.5%).
But it is “just ok” if all of the other guys were at 99.8%. How well he does has to be measured against the other guys.
I didn’t bother to check all the Rasmsussen races, but I believe you and I’m not surprised by that either. I’d hope that anyone who aggregates polls would be more accurate than virtually all individual pollsters. Just curious, do you know how RealClearPolitics averages did in 2010?
OMG! An Asian Dad Conservative!
And, in Rasmussen’s defene, they were apparently the best individual pollster in 2008: http://www.fordham.edu/images/academics/graduate_schools/gsas/elections_and_campaign_/poll%20accuracy%20in%20the%202008%20presidential%20election.pdf
Well that was accuracy based on the final poll only. I wonder if anyone did an analysis to see who had the most consistent as well as accurate polling- because IIRC, in 2008, Rasmussen was off by quite a bit for most of the race between McCain and Obama, and at the end it “suddenly” became accurate.
As Silver has repeatedly noted and factored into his model.
Do you have a link? Not that I don’t believe you, I just didn’t know that and would like to read about it.
iiandyiiii This is for 2010, not 2008, but you might be interested. Nate Silver: For Rasmussen, however, the Republican lean in its polls ran pretty much wire to wire. It had a significant Republican house effect early in the election cycle and a significant Republican house effect late in the election, and it would up turning into a significant Republican bias on Election Day. When 'House Effects' Become 'Bias' - The New York Times
That was a preliminary report. The author updated his work after final election numbers came in and Rasmussen fell to the median in the final rankings.
I thought about that, but, in the end, I determined that it was irrelevant. Let’s assume Rasmussen is the best pollster. Rasmussen certainly believes that it is or else, why be in the pollster business?
So you have a choice between Rasmussen, the best and most accurate, or an average (or conglomeration) of a bunch of polls, all of whom would, by definition, have to be inferior to the best.
So if one poll is THE most accurate, an aggregate of all polls would only serve to drive the accuracy down as compared to the best.
Or, to flip it, I’d ask the question of why Nate Silver would bother to do all the work he does if there’s one pollster who’s more accurate than he could be. Why not just have a blog that says “here’s Rasmussen’s numbers. Enjoy!”
Speaking of Rasmussen, their poll today shows Obama and Romney tied at 48.