Michelle Obama: name Hussein is 'the fear bomb' - Page 3

Originally Posted by franksargent

Actually, I believe either would beat McCain handily. It's rather obvious that the Democratic Party is highly energized this time.

And if the economy does take a true turn for the worst, like it did for 41, then watch out, it will be the death knell for the Republicans, and a certain landslide victory for the Democrats.

It's the economy, stupid



Sound familiar?

Originally Posted by ShawnJ

I'm left wondering after all of this whether we'd be better off without the 24/7 cable news channels.

Originally Posted by dmz

Damn, franksargent, how much worse do you want it? I've had enough bad news for the next ten years -- this housing market thing is tricky though -- it will be tough to sort out the all the pimps, johns, and whores before we can blame anyone.

I think you're wrong on HRC taking McCain, the polls I've seen:

Obama vs. McCain\tObama 47.4\tMcCain 43.5
Clinton vs. McCain\tClinton 45.5\tMcCain 46.3

Originally Posted by franksargent

Sample sizes are too small, and there are significantly large and diverse demographic groups, that the "random sample" assumption, isn't applicable to the actual population that votes on election day.

The truth is, that none of the polling groups, rigorously check actual voting data against polling data (and then place it into the public domain for others to judge the methods used to arrive at the assumed "randon sample," D'oh!)

For example, ±2σ ~ 95% confidence interval, meaning that if it is a truly random sample of the actual voting population, then we would expect, on average, the polling data and voting data agree 19 times out of 20, within the stated MOE. And you know what, I don't think I've ever seen any polling group, ever remotely come close to a 0.950 batting average for voting populations.

Originally Posted by BRussell

franksargent, maybe you know this, but polling groups don't do random samples, they do models based on who they believe will vote. And they are terrific when it comes to predicting general election voting a few days before the actual vote. They haven't been so good at predicting the primaries, because they don't have as much experience modeling them in all the different states.

But the real issue for the general election is that it's 9 months away, and the campaign hasn't even started. I'd trust someone's opinion who has a good understanding of politics and the candidates' strengths and weaknesses and the way the campaign will likely go over a poll trying to predict what will happen in 9 months. Predicting what will happen during the campaign is a much bigger issue than the sampling accuracy of the polls.

Originally Posted by mydo

Don't they always base the sample on "likely voters"? Depending on what you want to call "random" BRussell isn't incorrect.

Statistics is a messy nasty game. Surveys of people are almost useless IMO unless the tip 1/3 to 2/3.

Originally Posted by franksargent

... they always take a sample from the general population, always. I've never seen a poll where the sample size, N, wasn't mentioned. Now if they take additional information demographics, then they can proceed to making various modeling assumptions as to the assumed voter turnout on election day.

Now a national poll usually has 2,000 < N < 4,000, at which point all polling groups are polling for essentially the two leading candidates, but throw in a Nader or a Perot, that tends to muddy the waters somewhat. So we have a dozen (or more) groups taking the aforementioned surveys, what, a dozen times (or more) over a six month period?

These numerous polls if they show similar numbers and vary little (or consistently) over the time to election, then, and only then, do I start to get a warm and fuzzy feeling for the actual outcome.

But even then there is the EC to deal with, so that we would still need to poll all states (and DC) populations, at least for the states expected to be close (states where the 2*MOE is greater than the spread),

Originally Posted by franksargent

Actually, I believe either would beat McCain handily. It's rather obvious that the Democratic Party is highly energized this time.

And if the economy does take a true turn for the worst, like it did for 41, then watch out, it will be the death knell for the Republicans, and a certain landslide victory for the Democrats.

It's the economy, stupid

Sound familiar?

Originally Posted by BRussell

I'm not sure what you mean by "they always take a sample from the general population." Who wouldn't be the general population?

My understanding is that national polling groups have long since given up on trying to get random samples, and instead try to get certain percentages of certain types of people (men ,women, repubs, dems, blacks, whites, young, old, etc.), and within those categories the individuals are as random as possible, i.e., stratified random sampling. Right now, for example, they're trying to figure out if Obama is going to get a higher black turnout in Texas to know if they should include a higher percentage of blacks in their model - they don't just randomly call people and go with however many blacks they happen to get on the phone. They go for certain percentages based on previous data. Each polling group has their own proprietary model, and which model they use will go a long way in determining their accuracy.

Even though dumbass news orgs love to talk about the margin of error as if they've become mathematicians, the models used (and the question wording and such) are much more important for accuracy than the confidence intervals they report.

BTW, I've never seen a political opinion poll with an N between 2000 and 4000. It's usually around 1000 and often significantly less which, stratified sampling or no, is plenty big enough to estimate 100 million voters - central limit theorem and all.

Methodology

The Harris Poll® was conducted by telephone within the United States between February and November 2006 among a nationwide cross section of 10,032 adults (aged 18 and over). Figures for age, sex, race/ethnicity, education, region, number of adults in the household, size of place (urbanicity) and number of phone lines in the household were weighted where necessary to bring them into line with their actual proportions in the population.

All surveys are subject to several sources of error. These include: sampling error (because only a sample of a population is interviewed); measurement error due to question wording and/or question order, deliberately or unintentionally inaccurate responses, nonresponse (including refusals), interviewer effects (when live interviewers are used) and weighting.

With one exception (sampling error) the magnitude of the errors that result cannot be estimated. There is, therefore, no way to calculate a finite "margin of error" for any survey and the use of these words should be avoided.

With pure probability samples, with 100 percent response rates, it is possible to calculate the probability that the sampling error (but not other sources of error) is not greater than some number. With a pure probability sample of 10,032, one could say with a ninety-five percent probability that the overall results would have a sampling error of +/-1 percentage points. However that does not take other sources of error into account.

These statements conform to the principles of disclosure of the National Council on Public Polls.

Originally Posted by BRussell

frank if that supports your position, I guess I'm not sure what your position is. I'm just saying that they rely heavily on stratified sampling for election polling rather than simple random sampling. This was a big issue that everyone was talking about in the last few election cycles because different agencies used different models for weighting how many Rs and Ds would get out to vote, and polls that used different models tended to have different results.

BTW, that 10,000 number is absurd - that's a combination of all of that group's polls in that year, not a single poll, which tend to have 500-1000 respondents. And I don't believe that link applies to election prediction polling, it applies to estimating the number of people who call themselves Rs and Ds, which is a different issue. When they're trying to predict votes, they rely on their voter turnout models; when they're trying to estimate opinions of people in the country in general, I assume they only weight by census data like you said.

Originally Posted by Flat Stanley

Wing nut radio is encouraging Republicans to vote for Hillary today so that the campaign will continue and she will continue to sling mud at Obama. They are very afraid of Obama.

Originally Posted by BRussell

frank: Sample sizes are most definitely not the problem with election polling, especially polling McCain vs. Obama or Clinton. The main problem is that it's 9 months before the vote and the campaign hasn't even started.

But have fun.

Originally Posted by MaxParrish

True. The conservative leaders of talk radio do want Hillary to keep up the fight. It helps divide the Democrats and assures us of good entertainment. And they'd like to think it is working...

LATEST RCP POLL shows Hillary ahead in Texas by 1.7% average.

GO HILLARY RODHAM CLINTON -

Originally Posted by franksargent

So there's no such thing as sampling errors or sampling biases?

Saying something isn't so, because "it isn't so," does not appear to me to be a sufficient answer. \

Originally Posted by mydo

I would guess that the "weighting" is the tricky part.

Originally Posted by BRussell

Right, that's exactly what I said, that there's no such thing as sampling error or sampling bias. You got it perfectly. I didn't say "sample sizes aren't the problem" or "the problem is that the election is 9 months away."

1) Look, if they could poll every single person in the US about who they're voting for in November, it wouldn't be any more accurate than the existing 800-person polls, when it comes to predicting the outcome of the election. That's because things will happen during the campaign that will move significant numbers of voters from where they are today.

2) It's one of the common myths about polling that they need many more people in their sample than they commonly use because the population of all US voters is so large. The central limit theorem shows why several hundred is just fine whether your population is a few thousand, a few million, or a few hundred million. That's why I addressed this issue of sample size.

3) On the models they use: Voters are not a random sample of Americans. Certain people are more likely to vote than others - older, whiter, wealthier, etc. That's why their likely voter models are so much more important than their sample sizes or their confidence intervals. I'm not saying "sampling error doesn't exist" - I'm saying that their sample sizes are large enough, and the real error comes from the models they use, which are mostly guesswork based on human experience and judgment: "I think we should increase our estimate of black turnout by, um, 4%! because of Obama being a black candidate."

Here's some interesting discussion of election polling that I followed regularly (when it was active), about likely voter models and where error comes from in polling.

Anyway, I'm sorry to have challenged you on the issue, I didn't want to get into a pissing match about it. I just thought I could throw in some useful info about the issue because it's something I had researched in the past. I guess I forgot this is the internet...

Originally Posted by ShawnJ

Never argue with a retired guy with an internet connection.

Originally Posted by BRussell

Alright man, you win.

Originally Posted by FormerLurker

Close!


It's the Hardest Part.


Just ask Tom Petty!

Originally Posted by Frank777

I'm surprised that no-one's mentioned the tempest over the Clinton camp's latest firebomb. (Or maybe I've missed it in another thread?)

Is this likely to hurt Obama or Hillary more?
Is Hillary really being persecuted by a vast sexist media conspiracy?

Originally Posted by Frank777

I'm surprised that no-one's mentioned the tempest over the Clinton camp's latest firebomb. (Or maybe I've missed it in another thread?)

Is this likely to hurt Obama or Hillary more?
Is Hillary really being persecuted by a vast sexist media conspiracy?

Mondale's campaign was already far behind the Republican ticket when Ferraro joined the ticket, and one issue that hurt her credibility was her disclosure of her husband's tax returns. In July 1984, she said she would release both her and her husband's tax returns. Yet a month later she backtracked and said she would release only her returns. Then she backtracked again, saying her husband would release "a financial a tax statement" on August 20. But she must not have consulted her husband, because Zaccaro initially refused.[7] To her astonishment news surfaced that when she was a baby both parents had been under federal criminal indictment for gambling; the charges were dropped when her father died. After the election the House Ethics Committee officially criticized her mishandling of campaign finances.[8]

Originally Posted by mydo

I heard her on NRP a week or so ago saying largely the same thing. She may be right.

Originally Posted by Frank777

I'm surprised that no-one's mentioned the tempest over the Clinton camp's latest firebomb. (Or maybe I've missed it in another thread?)

Is this likely to hurt Obama or Hillary more?
Is Hillary really being persecuted by a vast sexist media conspiracy?

Mrs Ferraro, the Democratic vice-presidential candidate in 1984, said that Mr Obama was only where he was in the race because he was black. Race being as ticklish a subject as it is in America, most politicians would retreat and praise Mr Obama when such comments caused a stir. Instead Mrs Ferraro said that the outcry made her a victim of reverse racism. A spokesman for Mrs Clinton has said only that she disagrees with Mrs Ferraros comments. The Obama team asks why, if Ms Power had to quit, Mrs Clinton does not disown Mrs Ferraro. And if she doesnt, Mr Obamas supporters may lash out in frustration.

A campaign that degenerates into name-calling and mud-slinging will hurt Mr Obama more than it does Mrs Clinton. He has campaigned on messages of change and hope so he faces an unenviable choice in the long run-up to Pennsylvania. If he lets the Clinton team fling the brickbats without retaliation she may set the tone of the campaign. But respond in kind and his message of a new politics is tarnished. Even though he is behind there in the polls, Pennsylvania cannot come soon enough for Mr Obama.