WombatControl: BlueJay206: So CNN shows a tied race, while sampling D+11

this is obviously bad news...for Obama

Exactly.

So Silver takes a set of polls he admits have generally leaned towards Obama, looks only at the toplines, and concludes Obama is a favorite to win.

quantum_csc: whidbey: He's not going to win. He can't, damn it. I WANT a job in an oil field ferchrissakes...

Don't you mean coal mine? Oil production has gone up in the US under Obama.

No no, come on. Romey's going to expand that shiat. I'm talking about REAL jobs, not crap in a mine.

coeyagi: protectyourlimbs: Watching FOX news at work has me looking for my tinfoil hat, seems like they are sure Mitt is going to pull out a victory and the only thing I can see from all the real polls would be actual voter tampering...

WALNUT and PECAN had better go into hiding now. O'Queef will be gunning for them on 11/7.

Yeah, if Obama wins the electoral but not the popular this could have extra entertainment value...

 

Lets send it to overtime...

DamnYankees: WombatControl: That scenario is much more likely than Silver thinks.

And you say this based on...what?

Early voting figures for one.

Obama needs certain margins with early voters in key states to win - even in 2008 Election Day voters leaned Republican. The President is not pulling those margins right now. What we are seeing consistently is less early voters in Obama's column in key states like Iowa, Virginia, Colorado, and Ohio.

All the polls assume a 2008-level turnout for Obama or bigger. But we can test that hypothesis against actual numbers now. And the actual numbers don't support that.

Now, Obama could lose a good chunk of how 2008 support and still win. If it's a D+4 race instead of a D+7 race, Obama still wins. But if it's so much as an R+1 race, Obama is less than a 50% favorite in the Electoral College - by Silver's own data.

R+1 is well within the historical error range of national polls.

WombatControl: All the polls assume a 2008-level turnout for Obama or bigger. But we can test that hypothesis against actual numbers now. And the actual numbers don't support that.

Really? So when a pollster calls a random sample set, and it records R+4, they throw away that idea because of 2008 and add D+7? And you are the only person in the world that knows better?

Seems legit.

WombatControl: R+1 is well within the historical error range of national polls.

Why do you want Romney to win so bad?

Rush wants the poll samples to take the 2010 mid terms turnout into account.

Yea, ok.

DamnYankees: WombatControl: All the polls assume a 2008-level turnout for Obama or bigger.

Ah, I understand now. You're Dean Chambers.

Dean Chambers has the right idea, he just takes it WAAAAAY too far.

For one, "re-weighting" the polls doesn't work except in the roughest sense. You'd have to redo the samples. So you can "correct" for a skew by shifting the numbers around, but that's never going to be an exact measure. I do that informally to get a sense of where the polls might be given certain turnout assumptions, but that's unscientific.

Where Chambers really goes bugnuts is that he reweighs the polls twice. First on partisan ID, then on ideological ID. That's just not a valid measure, so it utterly destroys any validity to his model.

And further, Nate Silver does the exact same thing with 538. His model reweighs the polls too. He just does it based on a much larger set of factors - which introduces more opportunities for error to slip in.

So while Chambers is a nutjob, his first analytical step is at least partially valid. And yes, that is damnation by faint praise...

WombatControl: given certain turnout assumptions, but that's unscientific.

Where Chambers really goes bugnuts is that he reweighs the polls twice. First on partisan ID, then on ideological ID. That's just not a valid measure, so it utterly destroys any validity to his model.

And further, Nate Silver does the exact same thing with 538. His model reweighs the polls too. He just does it based on a much larger set of factors - which introduces more opportunities for error to slip in.

So while Chambers is a nutjob, his first analytical step is at least partially valid. And yes, that is damnation by faint praise...

Don't ever change.

WombatControl: Dean Chambers has the right idea

WombatControl: And further, Nate Silver does the exact same thing with 538. His model reweighs the polls too. He just does it based on a much larger set of factors - which introduces more opportunities for error to slip in.

In 2008, he correctly predicted the outcome of the presidential election in 49 of 50 states, and accurately identified the winner in all 35 Senate races.

I guess the errors just didn't take the opportunity to slip in.

NateGrey: WombatControl: All the polls assume a 2008-level turnout for Obama or bigger. But we can test that hypothesis against actual numbers now. And the actual numbers don't support that.

Really? So when a pollster calls a random sample set, and it records R+4, they throw away that idea because of 2008 and add D+7? And you are the only person in the world that knows better?

Seems legit.

Your example is crude, but yes. If you poll New Jersey and you end up with an R+4 sample your poll sampling is bad. If you poll South Dakota and your sample is D+5, your poll sample is bad.

So either you get a bigger sample (which is expensive and time consuming) or you reweigh the polls with an adjustment factor up or down.

The problem is that the pollsters are assuming a turnout model like 2008 - which is not a safe assumption. (PPP does this, for example.)

The other problem is that some pollsters refuse to weigh polls - but that assumes that both parties are equally likely to answer a pollster's call. That's not always true, for many reasons. (This is called the "shy Tory" effect in polling lingo.) If that's not true, then the sample is biased.

If the polls are wrong, it may very well be due to one or both of those endemic problems with public polling. Some elections are more difficult than others - 2008 was easy to predict with polling. 1980 or 2000 were not. 2012 looks to be more like the latter than the former.

WombatControl: NateGrey: WombatControl: All the polls assume a 2008-level turnout for Obama or bigger. But we can test that hypothesis against actual numbers now. And the actual numbers don't support that.

Really? So when a pollster calls a random sample set, and it records R+4, they throw away that idea because of 2008 and add D+7? And you are the only person in the world that knows better?

Seems legit.

Your example is crude, but yes. If you poll New Jersey and you end up with an R+4 sample your poll sampling is bad. If you poll South Dakota and your sample is D+5, your poll sample is bad.

So either you get a bigger sample (which is expensive and time consuming) or you reweigh the polls with an adjustment factor up or down.

The problem is that the pollsters are assuming a turnout model like 2008 - which is not a safe assumption. (PPP does this, for example.)

The other problem is that some pollsters refuse to weigh polls - but that assumes that both parties are equally likely to answer a pollster's call. That's not always true, for many reasons. (This is called the "shy Tory" effect in polling lingo.) If that's not true, then the sample is biased.

If the polls are wrong, it may very well be due to one or both of those endemic problems with public polling. Some elections are more difficult than others - 2008 was easy to predict with polling. 1980 or 2000 were not. 2012 looks to be more like the latter than the former.

Based on...what, exactly? You still haven't said anything substantive about any specific poll.

WombatControl: If the polls are wrong, it may very well be due to one or both of those endemic problems with public polling. Some elections are more difficult than others - 2008 was easy to predict with polling. 1980 or 2000 were not. 2012 looks to be more like the latter than the former.

You base your idea that 2021 will be difficult to predict with polling solely on your belief (hope) that the polling data is wrong. Now THAT is some solid methodology.

magusdevil: WombatControl: And further, Nate Silver does the exact same thing with 538. His model reweighs the polls too. He just does it based on a much larger set of factors - which introduces more opportunities for error to slip in.

In 2008, he correctly predicted the outcome of the presidential election in 49 of 50 states, and accurately identified the winner in all 35 Senate races.

I guess the errors just didn't take the opportunity to slip in.

And again, for the millionth time predicting the outcome in 2008 was painfully easy - Obama was ahead by nearly 8 points. The only state that was questionable was Indiana - which is the state that Silver got wrong.

Most political predictions are easy. You would have come to the exact same conclusion as Silver if all you had done was used the RCP averages - or a pocket calculator.

Crowing about getting 2008 right is meaningless because 2008 was a blowout election from September on.

That argument for Silver being a political genius does not hold water.

magusdevil: WombatControl: If the polls are wrong, it may very well be due to one or both of those endemic problems with public polling. Some elections are more difficult than others - 2008 was easy to predict with polling. 1980 or 2000 were not. 2012 looks to be more like the latter than the former.

You base your idea that 2021 will be difficult to predict with polling solely on your belief (hope) that the polling data is wrong. Now THAT is some solid methodology.

It's going to be really difficult to predict considering we don't know who the candidates are yet.

WombatControl: DamnYankees: WombatControl: That scenario is much more likely than Silver thinks.

Fine. Great. Listen, make your prediction. You think Romney is going to win, really? Then nut up and say so. You think the majority of Electoral Votes are going to Romney, then state your number. Will Romney get 271? 290? 303? Make your prediction and I'll make a note in the election night thread, along with all the predictions from the people who are not hiding behind a fake name on the internet. Proceed, Wombat. Then, the glory will be all yours if Romney pulls off a win. Go for it.

WombatControl: magusdevil: WombatControl: And further, Nate Silver does the exact same thing with 538. His model reweighs the polls too. He just does it based on a much larger set of factors - which introduces more opportunities for error to slip in.

In 2008, he correctly predicted the outcome of the presidential election in 49 of 50 states, and accurately identified the winner in all 35 Senate races.

I guess the errors just didn't take the opportunity to slip in.

And again, for the millionth time predicting the outcome in 2008 was painfully easy - Obama was ahead by nearly 8 points. The only state that was questionable was Indiana - which is the state that Silver got wrong.

Most political predictions are easy. You would have come to the exact same conclusion as Silver if all you had done was used the RCP averages - or a pocket calculator.

Crowing about getting 2008 right is meaningless because 2008 was a blowout election from September on.

That argument for Silver being a political genius does not hold water.

He's been closer on a lot of things than the average, before and after 2008. I don't think anyone thinks him a political genius, but I do think he is pretty much a statistical/math genius (at least compared to the vast majority of people).

What's pretty great about Nate Silver is that he is solid with his numbers while also being unusually good and communicating his ideas and findings to a larger audience. You can find plenty of good communicators, and plenty of amazing statisticians, but having both in one person is rare and pretty handy.

DamnYankees: WombatControl: predicting the outcome in 2008 was painfully easy - Obama was ahead by nearly 8 points.

It's amazing then how zero pundits (according to this aggregation) managed to exactly predict the EC outcome, and only 1 of them managed to come closer than Silver.

http://www.huffingtonpost.com/2008/11/02/election-predictions-pund_n_ 1 40149.html

Oh, thank the gods. Has Wombat made his prediction?

WombatControl: All the polls assume a 2008-level turnout for Obama or bigger. But we can test that hypothesis against actual numbers now. And the actual numbers don't support that.

All the polls assume that people will actually be allowed to vote. But the Republican suppress the vote efforts are working so the numbers are much lower than 2008. People having to stand in line for over 8 hours to cast their votes, that's the new reality, and guys like WombatControl couldn't be prouder.

Nadie_AZ: Wow. One guy has totally owned this thread. Well done.

If by that you mean "left after he was asked to put up or shut up"

DamnYankees: http://www.huffingtonpost.com/2008/11/02/election-predictions-pund_n_ 1 40149.html

Because there's a difference between pundits and statisticians.

Go look at the 2008 RCP averages - if all you did is go by the averages of the polls and nothing else, you would have predicted 49 out of 50 states, except for Indiana. Which is exactly what Silver predicted.

This article does a good job of explaining why Silver's much-vaunted model isn't that great:

This is true, and it leaves us with only the question of what information Silver's model may actually be adding to the field of alternatives. The answer could conceivably be "Less than none", if his model (or his style of model-building) is inherently prone to getting the easy calls right and blowing up completely in the face of more difficult ones. (Taraji P. Henson Alert!) It is worth pointing out that a couple of statisticians have given us a potential presidential equivalent of the Marcels-a super-simple model that nailed the electoral vote the last two times (and that actually is fully specified).

DamnYankees: WombatControl: This article does a good job of explaining why Silver's much-vaunted model isn't that great:

Silver's model isn't that great - his writing is. Other models (like the Princeton Election Consortium) are more accurate, and they are even more bullish on Obama.

I'll give him that - his writing is very good. I read him daily myself, and there's a lot of things I agree with him on.

All the models based on polling are assuming that public opinion polling is still highly correlated to the actual election results. Most election cycles that's a perfectly safe assumption. But sometimes they are not - 1980 comes to mind. And if the public opinion polls (especially the swing state polls) are off, then the models are off too.

And in a close election, they don't have to be off by very much at all to have a huge impact. If there's as little as a 2% bias in the polls for Obama that's consistent throughout - then key swing states like Ohio go the other way.

WombatControl: Go look at the 2008 RCP averages - if all you did is go by the averages of the polls and nothing else, you would have predicted 49 out of 50 states, except for Indiana. Which is exactly what Silver predicted.

So then why isn't the RCP average much different from 538 or Sam Wang's prediction?

WombatControl: And if the public opinion polls (especially the swing state polls) are off, then the models are off too.

So then why was the polling not off in 2004 and 2008 but very certainly (according to you) off by a lot in 2012? What changed in specific polls?

cameroncrazy1984: WombatControl: And if the public opinion polls (especially the swing state polls) are off, then the models are off too.

So then why was the polling not off in 2004 and 2008 but very certainly (according to you) off by a lot in 2012? What changed in specific polls?

That would be the $64,000 question... and if I had the answer to that I'd be calling up a literary agent.

My guess is that polling has become less frequent, more politicized, and more difficult, and that trend has been going on for a while, but the 2008 election hid the true extent of it. 2008 is an outlier election because you had such a clear outcome and a massive amount of support for Obama and very little for McCain.

Or it might just be that the pollsters are doing what they normally should do and using the last election as a baseline for the first, which is leading them astray because 2008 was such a wave year.

The problem with this is that we can only guess what the answer may be until the numbers come in and we can start digesting them. It's entirely possible that the polls are dead-on right this year and there's just some odd data floating around. But the evidence points to more than that going on...

WombatControl: That would be the $64,000 question... and if I had the answer to that I'd be calling up a literary agent.

So you're absolutely sure that polling is less accurate than in 2008 but you have no idea how or why?

Good luck with that.

WombatControl: My guess is that polling has become less frequent, more politicized, and more difficult,

I forgot to respond to this part - actually, polling has become more frequent.

WombatControl:

You arent saying anything unique. There are threads full of people like you.

You are basically a louder Freeper.

Show us what your derp model tells you, what is your prediction?

DamnYankees: FYI Gallup poll today shows Romney up 1. Last time they polled he was up 5.

At the risk of repeating myself - The ol' Rasmussen Gallup double back. Try to keep the GOP momentum going during the run up, but start changing your prediction a day or two before the event so you can claim you were "right" all along.

NateGrey: WombatControl:

You arent saying anything unique. There are threads full of people like you.

You are basically a louder Freeper.

Show us what your derp model tells you, what is your prediction?

Ok, I just forced myself to read those comments. And by the end I was laughing out loud at my desk. Those people are delusional. I got a real good chuckle from the guy that said "those commies at pew." By the end of that thread, I had fully decided that I will be on freerepublic for the duration of the election day along with having Fox News rolling in another window. The butthurt is going to be so epic.

cameroncrazy1984: WombatControl: That would be the $64,000 question... and if I had the answer to that I'd be calling up a literary agent.

So you're absolutely sure that polling is less accurate than in 2008 but you have no idea how or why?

Good luck with that.

There are a few reasons why:

1. Bimodal distributions in polling, when the results should be normally distributed.

2. Early voting not showing the big D turnout the polls are indicating.

3. Discrepancies in the state polls versus the national polls - which despite Silver's argument still exist.

I harp on the ABC poll, but it's a good example. They have a D+11 sample - which is patently ridiculous both on its face and based on hard data. Yet they show Romney winning independents by a whopping 22 points! A candidate that far up with independents should be running away with the election.

Now, maybe it's all statistical noise and the polls really are right. But arguing the polls are right base on the polls themselves is not a strong argument at all.

WombatControl: cameroncrazy1984: WombatControl: That would be the $64,000 question... and if I had the answer to that I'd be calling up a literary agent.

So you're absolutely sure that polling is less accurate than in 2008 but you have no idea how or why?

Good luck with that.

There are a few reasons why:

1. Bimodal distributions in polling, when the results should be normally distributed.

2. Early voting not showing the big D turnout the polls are indicating.

3. Discrepancies in the state polls versus the national polls - which despite Silver's argument still exist.

I harp on the ABC poll, but it's a good example. They have a D+11 sample - which is patently ridiculous both on its face and based on hard data. Yet they show Romney winning independents by a whopping 22 points! A candidate that far up with independents should be running away with the election.

Now, maybe it's all statistical noise and the polls really are right. But arguing the polls are right base on the polls themselves is not a strong argument at all.

You may want to get a sack of ice to drape your statisticals over tomorrow.

WombatControl: 1. Bimodal distributions in polling, when the results should be normally distributed.

Why should the results be normally distributed? Were they so in 2004 and 2008?

WombatControl: 2. Early voting not showing the big D turnout the polls are indicating.

How so? Dems have an advantage in Ohio and Va in early voting.

WombatControl: 3. Discrepancies in the state polls versus the national polls - which despite Silver's argument still exist.

How so? Rasmussen even has the national polls tied. How does that show a discrepancy? Romney could have a bigger lead in his "safe" states than Obama does. Doesn't mean there's a discrepancy. Please show your work.

cameroncrazy1984: WombatControl: 3. Discrepancies in the state polls versus the national polls - which despite Silver's argument still exist.

How so? Rasmussen even has the national polls tied. How does that show a discrepancy? Romney could have a bigger lead in his "safe" states than Obama does. Doesn't mean there's a discrepancy. Please show your work.

And even if there is a discrepancy, why should it be the national polls which are right and the state polls which are wrong instead of the reverse?

coeyagi: Reality for the next 36 hours, Farkers:

Every polling thread until 11 pm EST Nov. 6 will be hijacked by morons who dispute everything the EXPERTS say. If you were an expert, WC, you'd be getting paid for your "thoughts".

So everyone... Deal with it.

Ignore early, ignore often.

Actually, I am favoriting and coloring them according to their type. "Delusional Romney Supporter", "Paid GOP Shill", "Conservatroll" and a special one for "Channeling Phil Herup." That way, once the news hits, and the butthurt starts flowing, I can scroll quickly to where the hemmorrhaging is. I figure I wont have a ton of time to sit around soaking in the asshurt, so I should start preparing now to make best use of my time. I also have Mike_LowELL in bright green so I can get his take on things, because I know the parody of butthurt will be beautiful if he stays in character.