Election Prediction Poll...

[Ddrak]

So, just for kicks, a poll to predict the election.

[Garrdor]

Looks like 4 more years

[MeGusta]

I am predicting a Romney win contrary to polling based on a misrepresentation of party identification sampling.

[Partha]

I am predicting a Romney win contrary to polling based on a misrepresentation of party identification sampling.

Nice bridge.

[Harlowe]

http://fivethirtyeight.blogs.nytimes.com/

His chances have dropped down to 14.9% chance to win.

[Fallakin Kuvari]

I am predicting a Romney win contrary to polling based on a misrepresentation of party identification sampling.

Theres also the fact that Romney has been leading with independents in polls throughout October.

[Harlowe]

You're not going to be able to hide behind wishful thinking.

[Fallakin Kuvari]

You're can't hide behind Nate Silver's "math".

[Ddrak]

It's all chances. All outlets that work on an electoral college basis are showing a better chance for Obama than Romney, but polls are polls and have enough slack in them to make it far from a "done deal".

The only accurate polls are the ones that take a truly random sampling and just use that raw result. Anything else is blatant statistical falsehood.

Dd

[MeGusta]

^^This

[Lurker]

The only accurate polls are the ones that take a truly random sampling and just use that raw result. Anything else is blatant statistical falsehood.

So you don't think pollsters should weight results to match known demographic data? You think it provides an accurate picture to poll 2000 random people and report the results even if 1900 of them happened to be, through pure chance, old white men? Pollsters weight for known factors for a reason. Not for party ID, but for age, sex, race, so the random sample is representative of the population.

[Fallakin Kuvari]

The only accurate polls are the ones that take a truly random sampling and just use that raw result. Anything else is blatant statistical falsehood.

So you don't think pollsters should weight results to match known demographic data? You think it provides an accurate picture to poll 2000 random people and report the results even if 1900 of them happened to be, through pure chance, old white men? Pollsters weight for known factors for a reason. Not for party ID, but for age, sex, race, so the random sample is representative of the population.

Yea, because all the internals on these polls have been leaning heavily republican....

The internals on these polls paint a much different picture than the media is reporting; hell the movement on the ground paints a much different picture than what the media is reporting.

[Lurker]

Yea, because all the internals on these polls have been leaning heavily republican....

Fallakin, I was just giving an example in my question to Ddrak. I could have easily asked if the poll would be accurate if the random sample only contained single white women.

The internals on these polls paint a much different picture than the media is reporting; hell the movement on the ground paints a much different picture than what the media is reporting.

You can only make the first part of your statement by cherry picking internals for a few polls. The second part is just wishful thinking. But hey, we'll find out in 48 hours.

[Ddrak]

So you don't think pollsters should weight results to match known demographic data? You think it provides an accurate picture to poll 2000 random people and report the results even if 1900 of them happened to be, through pure chance, old white men? Pollsters weight for known factors for a reason. Not for party ID, but for age, sex, race, so the random sample is representative of the population.

Nope. If they are that worried about the poll from the demographics then they should throw the whole thing out and start over.

Let's take the 1900/2000 old-white-men argument. Let's assume the demographics say 300/2000 should be old-white-men. If you reweight then you've got an effective sample of maybe 130, with a 95% confidence around the +/- 9% which is entirely worthless in an election poll. If you take the much more likely case of a few percent out from demographics, I'm not sure the increased error from tinkering really does much to the result.

What you're advocating is actually doing a bunch of parallel polls for slices of the population and then combining them?

The real problem I have is weighting for something directly correlated with the result, particularly when you don't have stronger stats on that factor. Polls do weight for "likely voter party alignment" and the like, which is where I was really driving against.


As for what the polls are saying - all state-by-state poll tracking sites are showing a shift to Obama over the last few days. Take that however you want.

Dd

[Fallakin Kuvari]

The internals on these polls paint a much different picture than the media is reporting; hell the movement on the ground paints a much different picture than what the media is reporting.

You can only make the first part of your statement by cherry picking internals for a few polls. The second part is just wishful thinking. But hey, we'll find out in 48 hours.

Have you looked at the internals on any of these polls?

[Lurker]

The real problem I have is weighting for something directly correlated with the result, particularly when you don't have stronger stats on that factor. Polls do weight for "likely voter party alignment" and the like, which is where I was really driving against.

But we do have strong stats for the demographic makeup of the population. All pollsters make sure their samples are representative of that makeup. So if you want the most unaltered poll results you could look at the "registered voter" results, prior to the application of any "likely voter" screening. "Likely voter" screens are where the guesswork comes in. Don't conflate "likely voter" screens with weighting for demographics.

Have you looked at the internals on any of these polls?

I know from personal experience that if you have to start diving into poll internals for good news your candidate is not winning. I'd love to see your evidence though. Please post the internals for all the polls and why you think it indicates that Romney is actually ahead.

[Embar Angylwrath]

You're not going to be able to hide behind wishful thinking.

Like the Dems did when they thought they'd keep the House?

[Harlowe]

What would that have to do with anything? Or is that just a weak comeback? If people try to convince themselves of something they just wish would happen in spite of information to the contrary, they are just kidding themselves. Dem or Rep.

I don't think Romney has a chance of winning anymore. I did several weeks ago, it seemed 50/50 to me, but now I don't feel he really has a chance. If he does, it's the most cynical win ever and the voter suppression in which several state reps have said would deliver their state to Romney, paid off.

[Lurker]

I'm sure it wasn't his intention, but Embar's little dig actually bolstered your point.

[Harlowe]

Along with all the generally negative press for Romney (previous FEMA comments, Christie praising Obama, Bloomberg supporting Obama, tax loop holes, Chrysler coming out several times to calling him out on his ad, along with his and Ryan's reputation for not being very honest) and general good news for Obama (like jobs numbers, economy, gas prices, his response to Sandy) - I just don't see an electoral likihood of him winning - he's not going to get NH, WI, IA or even OH imo. He will most likely get FL. But even CO and VA lean Obama, so it's possible but not likely. Even with those that will only take him to 257. He would need to get all of those and either WI plus one more or all of those and OH.

Stranger things have happened, but Romney comes off as the most dishonest sales guy in the room and Ryan comes off as a weasle of epic proporations. Romney has most likely been tax evading into the tens of millions via Caymen's or the Irish-Dutch loop hole and by borrowing his church's exempt status http://www.bloomberg.com/news/2012-10-2 ... tions.html, so the guy isn't giving a trust me, I'm a decent guy vibe to the public anymore and he hasn't all along for women, hispanics or minorities.

The guy is losing.