This nerd pie fight is getting silly and reporting on it is even sillier.
1 Like
My view is that in this election you have polls all over the place and that all of them are right. The biggest difference in the polls is the assumption on who is going to vote. The real answer is nobody knows. This is different than passed elections.
There are other variables also in this election that make it different from other elections. After not knowing who will vote, Obamacareâs effect is the next biggest variable. Over half of those helped by Obamacare are White people with a high school diploma or less and living disproportionally in the South to include Kentucky, Gorgia, Arkanssas and North Carolina. Are these people really going to vote to have their health care taken away?
My own view is the Mitch McConnell who won in 2008 by 3% over an unkown and unfunded candidate winning half of the 23% of Kentuckians who have gained health care thanks to Obamacare will certianly lose regardless of what polls say. That is in the end whatever people say not enough of the half of Romneyâs 47% who the GOP spits on but in the past have voted GOP will do so this time for the GOP to win the Senate or have a good year. But then one has to wonder why half of Romneyâs 47% or half of the 23% of Kentuckains who McConnell at first tried to deny health care and now wants to take it away ever voted for the GOP anyway.
I still think thatâs incorrect. If you performed multiple trials, you can do some analysis showing that with X probability one model is better than another and itâs not just random chance. But in each trial you still have empirical results for which performed better that time. The statistical analysis is then performed based on the results of each trial. If you donât have individual results of which model performs better, you canât have statistical analysis of the entire data set to decide how statistically significant that performance was.
1 Like
Can we get Dean Chambers to weigh in?
Correct. And it should further be pointed out that he is citing his models documented superior performance to answer a claim that his model is inferior! Its performance in highly relevant when responding to this assertion.
2 Likes
Possibly; I myself doubt, though, that we really have enough data to establish which model is the superior one. If a model says that a particular candidate has a 60% chance of winning a particular election, then that candidate is going to lose in four out of ten scenarios. And if that doesnât happen over the long stretch, then the model is not calculating the odds properly.
Do you have any source for the âfailing badlyâ statement? Iâm not disputing what you say, the site is a massive disappointment. Just havenât been able to find much written on how itâs doing in the real world.
Me too, you can really see it if you read Wangâs whole response. Silver accuses Wang of a lot of things, and Wang rebuts every one of them.
"Silver claims: âIf youâd applied Wangâs technique in the past, you might still have had Republicans favored to win the Senate at this point in 2012 â even though the race had clearly broken toward Democrats by that time.â
Actually, using the âJune-to-nowâ method, at this point in 2012 the predicted probability of Democratic Senate control would have been about 70%. (A time-series of our 2012 Senate tracker is here, with previous published analysis here, here, and here.) "
His rebuttal is filled with these kinds of statements. Silver comes off as someone swinging in the dark. Iâm wondering if itâs his fame thatâs gotten the better of him, not paying as close attention to these polls now that heâs running his own news crew, or if 538 really is doing as poorly as I think they are. . .
1 Like
Iâve missed you 
Can we get Mollyâs Chambers to open up and provide input?
Of course you could do a statistical analysis. Take all the past races Nate called and all the races Wang called and compare them to the actual results. You have a table with 3 columns. Using ANOVA as the statistical test you can get a p value for Nate vs Wang. If it is >0.05, then you would have to say that they were indistinguishable. If p<0.05, then I would accept that one (presumably Wang) has been better at predicting so far (which doesnât mean they will be in the future).
If you visit Wangâs site, ask him to do such an analysis before he claims to be better.
Another whose Wang is biggerâŚopps I mean better, contest!
3 Likes
My hard-driveâs bigger than your hard-drive!
2 Likes
Sabermetric penis measuring. I love it.
Wangâs model may actually be better overall, but his explanation of 2014 appears flawed for a couple of reasons (for brevity, I cite the two most significant, as I see them on quick reflection):
-
As CO and IA have shifted Republican in the past few weeks, MI, KS, and MN have shifted away from the Republicans. So, CO and IA should effect the overall, but less than it seems they have. And before you say MN was never in doubt, it became even less in doubt. If Wang was weighting an IA vote 100 times greater than a MN vote in July or August, that judgment should be seriously questioned. While less recent, Oregon has absolutely shifted away from the Republicans since mid-summer as well.
-
And perhaps more importantly, Wang is wrong that where we are now is âpre-game.â To quote Wang:
Silver compares where we are in the race to the third quarter of a football game. Iâd argue that we are still in the pre-game period. The game - or more accurately, an election that will determine the direction of our national government - is still four weeks away. He and I arenât the players; weâre in the announcerâs booth. Meanwhile, on the field are the candidates, the strategists, and the men and women who are giving their time and money in key races. Theyâre the ones who are making the 2014 Senate campaign exciting.
Many states have already begun early voting. No excuse early voting states with Senate implications:
AK, AR, CO, GA, IA, KS, LA, MT, NC, OR SD, WV
So, for example, if some big scandal were to break regarding Sen. Merkley on 10/25, it may well be insufficient to change to outcome in OR. But even including only states where the outcome is generally agreed to be in doubt, voting has started or starts very soon in: AK, AR, CO, GA, IA, KS, LA, and NC.
Thatâs eight states in which every legal resident has the ability to vote early. That leaves only KY, MN (which I include generously at this point), and NH as states where, for all intents and purposes âgame dayâ truly is November 4. In all the others, a poll on 10/27 must be viewed in the context of predicted early vote, and thatâs problematic in a state like CO, where early voting continues to experience a long-term evolution. In other words, over the course of a two-day tracking poll, as much as ten percent of the vote (likely less, depending on the state) could be mailed, in transit, received, or tabulated.
Wangâs model may well be better, but some of his explanations for WHY itâs better need some improvement.
My Wang Has âMatched Or Outperformedâ Silverâs Since 2008?
Interesting use of term gaslighting in headline of this story on main page. Any comment on that, Josh or surrogate?
Would it be inappropriate to say Iâd rather have a good Wang than a little Silver?
Bingo.