Poll oversamples?
- Thread starter PriddyBoy
- Start date
It baffles me that after the Romney election, you all would still be sucked into this 'skewed'/'rigged' poll nonsense. Also, lmao that the reddit thread refers to 538 as a 'friendly' polling site. You all are too much. But whatever floats your boat.
You do realize that 'oversampling' polls is most likely referring to larger total sample sizes (which result in smaller margins of error) than actual targeted 'oversampling' of specific demographics.
I suppose I could open up a set of blueprints, without any engineering background whatsoever and offer a valuable opinion on the design and structural integrity of the new Freedom tower. I'll be sure to run my analysis by the OT board for vetting.
You do realize that 'oversampling' polls is most likely referring to larger total sample sizes (which result in smaller margins of error) than actual targeted 'oversampling' of specific demographics.
I suppose I could open up a set of blueprints, without any engineering background whatsoever and offer a valuable opinion on the design and structural integrity of the new Freedom tower. I'll be sure to run my analysis by the OT board for vetting.
Where to even begin? Probably because skewing/oversampling aren't even remotely the same thing? Because it serves no purpose to self-delude on polls? And ultimately, even if we assume that something nefarious was intended (no evidence for this) why would multiple independent and reputable polling companies, whose reputation depends on accuracy, not favorability towards particular candidates, go along with some plan to intentionally make their products less accurate?
Not one bit of it makes any sense within the context of 'rigging' or conspiratorial agendas. Oversampling to provide larger sample sizes and higher precision? Yeah, I could potentially buy that. But how could anyone answer the specifics from one email without additional context?
To me this is the ultimate disservice of the wikileaks stuff. Releasing stuff that has a direct bearing on the public good? Absolutely, I'm all for it. The Dept. of State stuff, info that gives insights to policy, etc.? Yes, it's important information to know. But the back-biting among staff? The emails about Bernie? Stuff like this where there isn't enough information to know what people are truly talking about, simply to troll low-information partisans? I tend to agree with Larry Leissig's take:
http://nymag.com/daily/intelligencer/2016/10/larry-lessig-has-a-very-good-attitude.html?mid=atlantic
Not one bit of it makes any sense within the context of 'rigging' or conspiratorial agendas. Oversampling to provide larger sample sizes and higher precision? Yeah, I could potentially buy that. But how could anyone answer the specifics from one email without additional context?
To me this is the ultimate disservice of the wikileaks stuff. Releasing stuff that has a direct bearing on the public good? Absolutely, I'm all for it. The Dept. of State stuff, info that gives insights to policy, etc.? Yes, it's important information to know. But the back-biting among staff? The emails about Bernie? Stuff like this where there isn't enough information to know what people are truly talking about, simply to troll low-information partisans? I tend to agree with Larry Leissig's take:
http://nymag.com/daily/intelligencer/2016/10/larry-lessig-has-a-very-good-attitude.html?mid=atlantic
It seems to me that from over a year ago the Clinton machine has attempted everything it can to confuse and mislead voters, and drive the public narrative. Trump being Trump has made their job way easier than it should have been, but from day one their campaign has been a slight of hand magic trick and people are falling for it hook line and sinker.
Podesta's email to me is pretty clear. Any advantage, no matter how slight it may be, with polling results helps to drive that public narrative.
You mean a candidate for POTUS and her team strategize how to hide/deflect her weak points and focus on positive things for her in an effort to win an election. The nerve of those people!
What is more plausible to believe?
1. Team Clinton has conspired to skew polls from multiple companies, to suppress negative coverage of wikileaks, to exagerrate Trump's negatives, to bring false accusers forward to impugn Trump, to foment lies about Melania's convention speech, orchestrate release of a 'rigged' video of Trump saying awful things about women, finding a past miss Universe contestant to concoct false stories, holding a gun to Trump's head to self-destruct with 3 am tweets, rigging election results via outright ballot fraud, and generally coordinating with international bankers and corporations to stop Trump using any means necessary.
2. A significant majority of the American people simply aren't interested in electing an abject know-nothing, reality show, tangerine-faced sleaze to be President of the United States.
Cool story either way.
1. Team Clinton has conspired to skew polls from multiple companies, to suppress negative coverage of wikileaks, to exagerrate Trump's negatives, to bring false accusers forward to impugn Trump, to foment lies about Melania's convention speech, orchestrate release of a 'rigged' video of Trump saying awful things about women, finding a past miss Universe contestant to concoct false stories, holding a gun to Trump's head to self-destruct with 3 am tweets, rigging election results via outright ballot fraud, and generally coordinating with international bankers and corporations to stop Trump using any means necessary.
2. A significant majority of the American people simply aren't interested in electing an abject know-nothing, reality show, tangerine-faced sleaze to be President of the United States.
Cool story either way.
Someone is attempting to mislead. Oversample, according to Pew, would include a disproportionate number of subcultures ( Black, Youth, Latinos) in the count. It is not simply enlarging the sample, but enlarging the subculture only. Most honest people would consider that is to add a bias to the sample.
Not sure about conspiring, but I take it a little like why they changed the reporting of poll results prior to the finish of the election in 04. The oversampling and/or skewed sampling would be a means to show Clinton in an overwhelming lead which will have an impact on a lot of the low information voters that just vote for who they think is either going to win or who is going to receive the most votes. Kind of like voting for the assumed winner because even not knowing, how can that many people be wrong? So just go with the flow. Lots of folks out there like that. They just hear so and so is not going to win, they don't bother to find out why, and just vote for the majority candidate.
Source? Story? No idea what you are talking about there.
Most of the polling right before that '04 election turned out to be very accurate. Liberals didn't want to believe it and trotted out all the reasons they thought it was 'skewed,' etc. But ultimately, Bush won a close race as most polls (and the averages) predicted.
What's more plausible to believe? The Clinton campaign is as dirty as many have suggested for years, and a large swath of the voting public is going to vote for either a liar and crook or an ego driven maniac.
It truly bothers me that people would even consider voting for Hillary or Trump considering what we know about them both now. There is ZERO defense of either one.
The reason is that one of them will win, period. For all the rhetoric about 3rd parties, and rainbow and unicorn-based optimism that one could actually win, Johnson never reached enough critical mass to be relevant. It's all well and good to feign outrage and say that people shouldn't vote for either option, but ultimately if one will win (which assuredly will happen) then the only recourse is to try and stop the worse of the 2. And that race isn't even close, even as bad as Hillary is. At least not for people who live on planet earth and deal in the realm of actual facts.
As someone who would actually like to see 2 viable parties for the sake of our country, it's pretty clear that Hillary can be easily beaten in 4 years and in the shorter-term, she comes without the boatloads of immediate risks that Trump brings to bear--i.e., his thin-skinned temperament (i.e., our national security and foreign relations), his obtuse rejection of facts and expert opinion, his pandering to know-nothing conspiracy theorists and alt media, his embrace of nativist/populist rhetoric, his childlike mentality that the only way he can lose is to be cheated, and his complete ignorance of the basic functioning of government that people take for granted.
Run Rubio over Clinton in 2020 and move on. Excise the nativist/populist cancer from what used to be a proud conservative party, and try to rescue credibility.
I don't believe in the "stop the worse of the 2" vote, it's what's wrong with our system.
I also certainly don't believe that the race isn't even close, and that Hillary is head and shoulders better than Trump. You can try and justify your vote all you want, but the woman is a liar, deceitful and as corrupt as they come.
I guess no one here has ever heard of weighted summation. Oversampling can be done within individual sectors, then weighted appropriately when added to the overall pool of participants to keep from skewing results. Pew discusses it on their web site. It's not difficult, and it reduces the errors for smaller sections of the population. The example Pew gave was a sample of 1000 people. If you go pro-rata, that includes 136 African American participants. The error on that sample is 10.5 points. If you increase the sample size of that sector of the population to 500, you reduce the error on that estimate to 5.5 points. You can do the math to scale the results from 500 down to 136 (reduce as a fraction, not by choosing which 136 you want to include.) It's a common polling method, and it improves accuracy instead of skewing results.
Oversampling is discussed about 2/3 the way down the page:
http://www.pewresearch.org/methodology/u-s-survey-research/sampling/
Oversampling is discussed about 2/3 the way down the page:
http://www.pewresearch.org/methodology/u-s-survey-research/sampling/
For partisans on a mission, your post and that link might as well be written in Yiddish. (And I'm not just talking about conservative partisans--liberal partisan group-think is just as bad).
I seem to recall, and I might be off base here, there was a lot of outrage about a couple of states being called early prior to polls closing and the argument being what i posted. So now, they won't call a state until the polls actually close in that state.
Sorry, as bad as he is, she is still far worse in my opinion. All of his shit is bluster and surface stuff that when actually doing a job potentially wouldn't be there. All of her stuff is actual shit she has done and will continue to do. You take any Hillary related conspiracy outside of the looney she killed Vince Foster type stuff and it's indicative of how she will continue to behave.
That happened in 2000, as the initial call of Florida for Gore came before some counties in the FL panhandle had closed the polls (part of the FL panhandle is in central time, not eastern). This was made worse by the fact that the call turned out to be inaccurate.
Methodology has been improved significantly since then, and it's now policy that no states will be called by VNS data prior to all polls closing in that state. In the vast majority of cases, it probably wouldn't matter, but I definitely agree in principle that people in line to vote should not be discouraged or potentially go home due to news breaking about the outcome in their state. Since 2000, there have been essentially no controversies about calls based on the revised methodology, which is used in concert with other live-fed information, polling data, and returns from bellwether counties by the AP and the various networks. Other than Karl Rove's little tantrum in 2012 about Ohio (which turned out to be baseless), there has been no significant on-air controversy about the states being called on election night. And when in doubt, the AP tally is considered to be the tally of record. Some races have gone into the wee hours (i.e., 2004 with Bush and Kerry waiting on Ohio), but to my knowledge there have not been any retractions by the AP or networks on air since 2000--at least not for states that were relevant to the final outcome; maybe some cases of states being called wrongly or revised after the overall race was already long since determined--i.e., moot point cases.
When you increased the black to 500, did you increase the others by the same percentage(proportionate), or did you freeze the total population at 1000.
I agree with the former 100% when you make the total sample size larger. If you are suggesting the latter, I call BS and you are polluting the sample
Was it 00? Ok, I thought for sure it was FL in 04 because I lived there at the time. Maybe I just watched a segment that night while they were discussing what happened in 2000. I just remember Susan Estridge and her nasally smoking voice yammering away about Kerry being the next President.
Regardless, and maybe it's not an apples to apples correlation, but the same logic kind of applies. If they are constantly hammering polling results showing the domination of Hildy in a 2 man race when the 4 person race shows a much closer comparison, people who might be coming out to vote for Trump, might just pack it in and stay home and not tarnish their soul by voting for what they see as the lesser of two evils. That was the point I was trying to make.
You can poll 500 people to give results that represent 136 people. If you read the description in my post, or better on the Pew site, you can increase the number for smaller percentages of the 1000 people, then reduce their impact to the overall results by multiplying by the correct fraction when combining to get the final aggregate results. It reduces the errors for the representative groups, and it improves the overall poll results. You can also increase the overall sample size to achieve similar results, but that's a more expensive proposition.
Of course you can get 500 to look exactly like 136. But, you damned surely cannot do it with random sample. You could get 500 to look like one, but that would require some selectivity. You are tipping the scales just a bit when you get a larger or smaller number. If you do what it appears you and Pew appear to be saying, you are really screwing up th philosophy that goes along with random sampling.
You didn't read the article or understand what was done. The sample was re-downsized/adjusted back downward for the final poll, but the initial over-sampling allowed for a more precise look at a particular demographic group.
This Tyler Durden guy is probably heavily invested in Aluminum.
http://www.zerohedge.com/news/2016-...em-playbook-rigging-polls-through-oversamples
http://www.zerohedge.com/news/2016-...em-playbook-rigging-polls-through-oversamples
How can you not do that with random sampling? You design the pool that you want, then randomly sample based on that. You can just as easily sample 500 people from a specific group randomly as you can 136. If you don't understand that, I suspect that you don't understand how random sampling works.
Please try to follow along. We've discussed what oversampling means, and it isn't a means by which polls are skewed. It is a means by which you can make your poll results more accurate by increasing some or all sample sizes. People who don't understand polling and the vocabulary of the field are over reacting to the use of a word that they don't know the meaning of. It's like someone hearing a holding call in football and expecting to see 2 players cuddling.
Follow along? I started this thread about this election with a revelation from the wikileaks Podesta emails. Call that into question or have your statistics 101 discussion, but don't change the topic and then tell me to follow along.
You seem to be allergic to facts. Go ahead and believe whatever you want about how the mainstream media is misleading everyone with bad polls. The intelligent ones in the room already know it's over.
What does that have to do with his response? Why do you ever imagine that you can speak for intelligent people? You are a dolt man.
I am not suggesting anything nefarious in oversampling for polls because I understand why pollsters do it. I do think you are overlooking the possibility that it can be done for less than scientific reasons as well. If someone wants a poll that says what they want to say they can massage the factual data with sampling. It only hurts their credibility.
That particular revelation is only a revelation if you don't know what oversampling means with respect to polling. We've discussed that in great detail with references to support oversampling as a legitimate and often-used tool. You follow up with the same revelation that you had to start the thread, more or less. That was my "follow along" comment. I'm not putting Podesta or Clinton or anyone in her campaign up for sainthood. I'm simply trying to discuss this particular issue and why it isn't of concern.
