Posted on 10/25/2016 10:46:41 PM PDT by Helicondelta
And the problem of all pollsters is when applying the weights acccording to the last election that may turn out to be correct or.......totally wrong.
There were about 125.9 million adult women in the United States in 2014. The number of men was 119.4 million. At age 85 and older, there were almost twice as many women as men (4 million vs. 2.1 million). People under 21 years of age made up over a quarter of the U.S. population (27.1%), and people age 65 and over made up one-seventh (14.5%).[13] The national median age was 36.8 years in 2009.[13]
About 2% more women than men so they overweighed women as well.
Correct. But they honestly have few other options
One way to correct the problem is to weight the same data differently and say
If the election is +7 D it will look like this
If the election is +4 D it will Look like this
If the election is +2 R it will look like this
But that would mean admitting that the pollster is, at some point just making an educated guess
BTW I think Trump takes about 53% of the vote or a good solid win. I hope I’m wrong and he gets more.
“A pollster can oversample on age, race, gender and income”
Saw a white paper the other day analyzing different poll methodologies in relation to the 2012 election. One of the conclusions:
“Demographic post-stratification, similar to that used in most academic and media polls, is inadequate”
It basically concludes that the LA Times/Daybreak methodology is the most accurate.
http://www.stat.columbia.edu/~gelman/research/unpublished/swing_voters.pdf
Doing two-candidate polls is stupid, when Gary Johnson is on the ballot in all fifty states plus D.C., and is polling (though probably won’t get) mid single digits.
If polling were the exact science that it is being so grossly misrepresented as, then why bother with holding elections at all? Just determine who wins based upon a poll. We're currently being told relentlessly that the election is over and Hillary has won, based upon nothing but minuscule samples of anywhere from a few hundred to a couple of thousand people which pollsters purport are representative of the intentions of over 100 million voters. Nonsense.
I think astrologers would be as accurate as pollsters. One simple question: If political polling has any validity at all, then how is it that various polls reach such wildly different conclusions? They love to cite the "margin of error" of their polls, as if the intentions of unpredictable humans can be measured with a micrometer. Well then, the margin of error for all of the polls as a group must be something like + or - 10 or 12% since that is about how far apart some of them are.
The polling scam is starting to closely resemble the climate modeling scam. Both are being used to drive an agenda and both are highly dependent upon the assumptions that are made, assumptions which we know are being manipulated in both cases to reach desired "results."
Bottom line: Ignore the phony polls and vote Trump. We ARE going to win.
I would say this even if they had Trump ahead. Registered Voter polls this late in the game are total BS.
GRAVIS last week had Trump -4 , so this is up 4 from last week.
Great NEWS!!
It’s just another way of saying Trump up 7. Of course, without saying Trump up 7. Which must never be said. Especially not on Her Thighness’ Birthday.
LOL. Pubbies refuse to play? Sorry, don’t cut it. Pure bovine scat. Thresholds for demographics means you ain’t done until the requirement is met. If it is easier for CNN to find libtards that is their fault. If they can’t find enough Pubbies to do a legit poll they need to either quit polling or change their bias.
Weighting polls is self selection. A truly random poll of 1012 voters without weighting is +/- 4% accurate.
Nationally, unless R stay home, the D v R is about 32 to 28. Any poll sampling about 4-5% is off on its sampling.
Again this would infer a tight/close race with a slight lead for Trump nationally.
If Trump is truly up only a few points nationally, the threat of a Trump win in popular vote but a Hillary Clinton EC vote is very real. Trump needs about 3-4% national lead to ensure that this scenario is not likely.
Assuming the monster vote is real, Trump has this handily... If it doesn’t and this poll is remotely accurate, could be a long night... And 2000 all over again.
Nah, true weight if a pure random sample of significant size is made should give you D+4 that’s about the registration difference between the parties.
R clearly have an enthusiasm advantage that may make Election Day Delta smaller than plus 4, but that is conjecture.
Not sure how they weighted this thing, but if it’s just no weighting adjustment and raw responses the 3 point D over sample would put trump up 2-3. If they reweighted the results before publishing to reflect a D+4 instead of their raw sample size then that’s not the case
Actually reweighting the samples to what you expect turnout to be is valid mathematically, but it does increase the margin of error. So this poll could be relighted as you suggest but for every deviation from the raw sampling you go, the MOE goes up.
Disappointing to see OAN publish this overweighted D+7 poll...I thought they were the honest news media.
They are showing what it looks like with the 2012 model and that’s very good news since the 2012 model will not be duplicated this year.
Trump will paint the US red...Hillary will be blue...
HAHAHAHAHAHAHA It’s 50/50 with a D+7 sample!!!! YES!!!
Can sample size eleventy billion and if it's RV instead of LV ... oh well ...
Looks like they added some Podesta sauce to make it look like it was tied.
Pray America wakes
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.