Posted on 08/14/2016 2:07:27 PM PDT by BigEdLB
Here is my analysis of oversampling during polling vs Gallups analysis of that actual party ID in a state is.
Thank you very much
Close. What it shows is the percent of oversample of Dems and Republicans VS what Gallup says are actual persons registered as D or R in each state......
Thanks.
Here is a link to the most recent Gallup Poll data on party affiliation:
http://www.gallup.com/poll/15370/party-affiliation.aspx
He has one that is state by state. that is the national one over time
bookmark
Thanks. Based on your calculations where you show 79 data points, 15 points with a Republican oversample, the likelihood of that kind of skewed result (or more skewed) based on approximation via normal distribution (Vassarstats.net) is:
< 0.000001 ie 1 in a million.
So either the pollsters are crooked or GOPers are more shy than Dems.
LoL People’s Press?
They should call themselves the Communist’s Press. That’s Pew Research once run by Madeleine Halfwit. You’re still plying your trade Zot Jamese777/Nero for Obama and Hillary.
Why, in 29 states where oversampling is recorded, 26 are oversampled to favor the ‘rats in the polls - only 3 are oversampled in favor of the ‘pubs - assuming a chance oversampling based on the random selection of participants, we would expect by chance that oversampling would favor the ‘pubs 14 times out of the 29 - instead we get only 3/29 in their favor - it’s been too long since I had statistics to remember how to figure the probability of this outcome, but it is quite small, probably well below the 5% level - we may thus reject as being too improbable the nulhypothesis that this sampling occurred randomly, and conclude that the pollsters are deliberately oversampling their polls to favor ‘rats - who said conservatives don’t like science......
This is my point. You explained it. Corrupt pollsters
First, this is a lot of work, so thanks for your efforts.
Second, in your opinion as someone who understands these things - do you think that most polling companies oversample Dems for the reason that people on FR would say - i.e. they want to skew the outcome to favor the Democrat as a way of influencing public opinion?
Or is there a statistically-valid reason for doing so?
In other words, does your chart help educate us on the inner workings of fair and honest polls, or does it expose manipulation?
Bottom line, any poll winds up with an assumption...either created or by happenstance of response....on the weight of an election. IF Dem turnout is TEN percentage points more than GOP, THEN perhaps Hillary is way ahead. But that has no basis in reality. As the OP stated, Gallup has party Id EVEN. Now pollsters could use other methods, but it is reasonable to state they Party ID question is even.
In 2012, If I recall, The Dems only had a 3 point advantage in Party ID polling at this time of the year. So from that score, the GOP is already polling 3 points better than 2012.
It is true this election is going to be bit odder than any since the Perot elections, because some of the particulars are all over the map...like Trump getting big Latino and Black support in some polls, and not in others. Like Dems voting Trump and GOP voting Hillary.
Then you have the Sanders voters, and Johnson polling above what I personally think his name recognition is.
So this election could be a real watershed true shift election...meaning....none of the usual models are going to work, and if they do, it would probably be dumb luck.
I see this as globalism vs nationalism, and Hillary’s corruption vs Trump’s perceived personality issues. That has the potential to provide a very unusual set of voting patterns in Nov.
And we haven’t even made it to the debates yet...and Trump has yet to even start a traditional ad/local office campaign.
Throw in the leaks, and any pollster who is telling you they have this nailed is lying.
The two-tailed probability of 3 or fewer out of 29 is: 0.000043 ie 43 in one million (Vassarstats.net).
Not very likely!
This is, I think what longroom was trying to say before he was mysteriously silenced. I use Gallup as my backup and compare their numbers to these other pollsters. No secret forula sauce. At some point, if it hasn’t already started, I expect some of these folks to start to make up stuff as they go along, if the surveys do match the narrative. Being a Democrat means never having to tell the truth.
I am in line with those here at FR who believe the Dem oversample is a way to create public opinion. I also think it is starting to backlash. Marist is th worst of the bunch, but PPP isn’t far behind.
“what longroom was trying to say before he was mysteriously silenced.”
You’re right. Longroom.com is gone.
You’re also right in that we have no guarantee that the polls we read about are truthfully reported. At all.
H3llary is losing.
Good post. I think you got almost all of it. One thing that one should add is that the response rate to any poll has declined over the years and is now (according to PEW) below 10%. (NB. Very rarely do the polling companies tell you how many interviews they have started, or how many phone numbers they have called).
As I have written elsewhere some research articles indicate that the errors in the polling have not increased despite the reduced response rate. That may well be true, but one has to be careful. Any election where there is more reason for voters for one candidate to not divulge their choice can cause very large discrepancies in the polling results.
Given the media’s relentless pursuit of Trump we may (cannot prove it, but we may) have such an election now. In that case the polls are not worth anything.
In November D+ will be 2+ or 3+ - it certainly won’t be 10+ or 16+
That’s wildly improbable. And even with the funny math, pollsters still can’t get Hillary over 50%.
Old rule of thumb an incumbent polling below 50% is usually going to lose.
CW though predicts a Hillary win in November so no pollster is going to go out on a limb and say the opposite.
So here we are.
Interesting to break it down by state like that. For instance, Florida shows polling data is 4.8% oversampled for Dems based on State party affiliation. That’s significant, unless by some deeper measure one assumes registered Dems are almost 5% more likely to actually vote in Florida, which would seem unlikely.
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.