still not how polling works
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
PoliGAF 2012 |OT4|: Your job is not to worry about 47% of these posts.
- Thread starter Amir0x
- Start date
- Status
still not how polling works
Father_Brain
Banned
Re: NH/PA, I have a very hard time believing that Obama could lose any state that Kerry won in '04, short of the Gallup +7 poll from yesterday actually proving correct.
I suppose this would be the only instance in history where Romney was against making money for himself?
I really would encourage people to stop dismissing polls which are not positive for Obama.
That does not mean you should automatically agree with them, either, a la a few posters here; the reasonable, logical approach is to factor them in to your analysis. Despite the ascension of Nate Silver, it seems many people still do not think probabilistically.
It is less likely that the Gallup/PPP/etc. polls indicate the current state of the race, but it is not impossible. My guess is that people would rather mock and dismiss them than factor them in because they find their conclusions depressing.
That does not mean you should automatically agree with them, either, a la a few posters here; the reasonable, logical approach is to factor them in to your analysis. Despite the ascension of Nate Silver, it seems many people still do not think probabilistically.
It is less likely that the Gallup/PPP/etc. polls indicate the current state of the race, but it is not impossible. My guess is that people would rather mock and dismiss them than factor them in because they find their conclusions depressing.
DragonSworne
Banned
Nobody is dismissing anything.
Shadow Hog
Member
A tie would be a Romney win, though, given the Republican House.
Indeed. If he can't get a mandate, I am almost certain that I would want the humorous fallout that would come with a split vote.
Manmademan
Member
This. There are a few pollsters with terrible track records, or none to speak of that get treated with a lot of skepticism, but no one here is saying Rasmussen etc should be dismissed entirely.
Gallup at +7 is definitely an outlier though
Didn't Nate already talk about this?
Really?
Every PPP poll thus far that has been good for Obama gets universal praise, then the NH one drops and people are like "umm we need more data, yep!"
What we have ONE poll of NH.
All the other state states at least 2 or 3n you know damn well how this works.
Yes.
There is a term for this in psychology called the "belief perseverance bias" which all humans do, in which they remember the things that agree with their mindset the most and ignore those that do not. This is just an example in here of what occurs every day in many situations. Your advice will not be headed by either side because it goes against how the human mind naturally works.
The Librarian
Banned
That could've been worded better.
TacticalFox88
Banned
Jesus, I swear this thread is reminding me of the old NPD days on gaming side except we're playing with polls instead of sales.
NPD doesn't include Walmart sales!
I'm white too, it's okay.
Hellsing321
Member
Panic confirmed!
Black Republican
Member
Frank the Great
Banned
right, which is why NH wouldn't be relevant in that scenario.
Yeah Romney is finish he went down 1.
The Librarian
Banned
That may be, but you should realize you're panicking so much that it's affecting the composition of your sentences.
...it's a seven day tracking poll. -2 would indicate a minimum difference between samples of 10 points.
Just from the past two pages, less than 24 hours:
Quite a few snarky posts in there that don't seem to be taking Gallup seriously. I can get more if you'd like.
PhoenixDark
Banned
There's nothing funny Bout Gallup. They're laying the foundation for voter fraud charges, if Obama wins
They simply need to be taken in context. As with all polls, they should be evaluated against the polling aggregates. If something is an outlier - Gallup - it's okay to call it such. But it should still be given appropriate weight in the same aggregates.
The Librarian
Banned
It's pretty clear Gallup's numbers are an outlier. I see it as more people making fun of Gallup than outright dismissing them.
I dismissed it because they are clearly an outlier when seen against every other reputable national poll. It was not a dismissal based on it being bad news for Obama. My response would be the same if Gallup was showing +7 Obama.
I absolutely agree. This is essentially a rephrasing of what you just quoted.
I am suggesting that many people are not doing that ("lolz Gallup," "fuck Gallup," etc.) and are simply dismissing it as an outlier to be ignored instead of an outlier to be factored in.
You don't make fun of outliers. That's not how statistical modeling works.
Right, because Gallup is at variance with all the other data. Obviously we factor it into the overall understanding of the situation (otherwise Obama would be obviously leading by a bunch), but the overall understanding of the situation suggests that Gallup is an outlier -- that they're not accurately representing things for whatever reason. Is it inappropriate to think this? I mean, here are another couple of "snarky posts that don't seem to be taking Gallup seriously:"
http://fivethirtyeight.blogs.nytime...-world/?partner=rss&emc=rss&smid=tw-thecaucus
http://www.tnr.com/blog/electionate/108816/whats-going-the-gallup-poll
Black Republican
Member
PPP: 49-48 to Romney in Iowa
This is not how aggregate analysis works. You don't decide the aggregate is right, and that outliers are to be ignored. You take all polls (conducted properly) seriously, and give greatest weight to the largest aggregate of polls.
I'm quite confident that Gallup's polling is one reason Nate Silver has Obama at ~2:1 odds instead of ~5:1, as he was polling less than a month ago. He is not ignoring or dismissing Gallup; he is clearly factoring them in and considering them seriously, despite the fact that they are currently outliers.
Why should they? They are so far and beyond the aggregate in any which direction.
They have been off in their prediction by more than their MoE all the way back to 1988:
They've been good about predicting the winner, but man, their margins are all kinds of wrong.
And not only that, but we have seen their cross-tabs, which completely overstate the south.
The worst part is that they have, by far, the largest polling sample. Why are they consistently so off?
Except I wasn't dismissing it as one to be ignored, in fact, I've been ridiculed in here numerous times because I cite them as a credible pollster, thus being the reason why I am so mad, because their huge gap in O/R LV will most certainly be factored in.
https://twitter.com/ppppolls/status/259340521795895297
You don't lead that largely in early voting and be behind 1 point.
worse than i expected.
wonder if there's a breakdown in the early voting.
yowza that's bad.
Technically, the longer Gallup keeps up being an outlier, the less impact that outlier status will have due to bias ratings, so the model IS slowly discounting them as being crazy over time.
- Status