• 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 2016 |OT11| Well this is exciting

Status
Not open for further replies.
No, I've understood you perfectly. Let me quote the precise part I mean:



This is wrong. You do not average polls because they use different methodologies. Some methodologies will be more accurate than others because the underlying assumptions are better. Some will be worse because the underlying assumptions are worse. All averaging a good set of assumptions with a bad set of assumptions does is produce a third result which is less accurate than the good set of assumptions. You would have been better off *not averaging* and simply using the good set of assumptions to begin with.

You average samples because it is effectively the same as creating a bigger sample - you know that on the balance of probabilities, you will be more accurate by doing so. Note that this isn't guaranteed; but expected in the strict mathematical sense of the word. There is a statistical proof of this. However, you shouldn't average methodologies - you should apply a single methodology, the one which has the most credibility. This is because there is no reason to suppose that on the balance of probabilities, averaging methodology will improve either the accuracy or the precision of your results compared to any given single methodology.

But we have no way of knowing which methodologies are more accurate and I am not confident that one methodology is best all of the time in all of the states. I am going to stand by my original thoughts on this which mirror Nate Cohn's.

You can see why we say it’s best to average polls, and to stop fretting so much about single polls.

You obviously know more about statistics than I do, so if you want a more fruitful discussion on why different LV methodologies are, in fact, a compelling reason to rely on polling averages take it up with Cohn on twitter.
 

Geg

Member
Wow, I'm gone for one day and a ton of good polls for Clinton come out.

I'm still too nervous to watch the debate I think. I'll just read about it later
 
D

Deleted member 231381

Unconfirmed Member
But we have no way of knowing which methodologies are more accurate

But this isn't true! We have all sorts of reasons. For example, we have strong reason to suppose landline polls won't produce an accurate sample because there are many eligible voters who don't have a landline.

You obviously know more about statistics than I do, so if you want a more fruitful discussion on why different LV methodologies are, in fact, a compelling reason to rely on polling averages take it up with Cohn on twitter.

I might, actually. I think that's a pretty terrible conclusion he's made there. It's the statistician's equivalent of "eh, fuck it, just fudge it a bit and it should turn out okay".
 
Oh, that Jill!

Stein had planned to challenge her exclusion from tonight’s event by hosting a rally outside the secured perimeter of the debate hall beginning around 5 p.m. ET. She said she will be live on Twitter’s Periscope app answering the same questions as Democratic nominee Hillary Clinton and Republican nominee Donald Trump on the debate stage.

Stein’s campaign said she will not “risk arrest” this time, because there is an outstanding warrant for her arrest over her involvement in a recent protest against a controversial pipeline project in North Dakota. Still, her campaign spokeswoman Meleiza Figueroa said they will attempt to get the “spirited demonstration … as close to the gates as possible."
 
But this isn't true! We have all sorts of reasons. For example, we have strong reason to suppose landline polls won't produce an accurate sample because there are many eligible voters who don't have a landline.



I might, actually. I think that's a pretty terrible conclusion he's made there. It's the statistician's equivalent of "eh, fuck it, just fudge it a bit and it should turn out okay".

You're right, FWIW. But...excluding the obvious, I think his point is that we don't know which voter turnout model is right and so we just have to average those (though I believe things like house effect should be included too).


Also 53Nate wrote something:
A good test of whether a model is too conservative, too aggressive or “just right” is whether it does a good job of matching new polls as they come out in a state. So far in this election, the FiveThirtyEight and Daily Kos Elections model — which also uses a trend-line adjustment — have done a good job of this, while other models sometimes lag behind the trend.

Silver again not understanding how this works...


Also, I think this is a big problem here:

FiveThirtyEight’s models also generally account for more uncertainty than other models — or at least they do in this election because the presence of a large number of undecided and third-party voters, who contribute to polling volatility. That helps Trump’s odds, since he’s (narrowly) the underdog in our forecast.

the model assumes uncertainty is better for the underdog...and while this might normally be true, this election is probably weird.

And if Clinton gains following the debates, FiveThirtyEight’s models will probably be among the quicker ones to detect it.

UGGGHHHH..he just earlier said PEC would be quicker since it's already assuming this revision.
 

Amir0x

Banned
I love flashback hour. God that 2012 first debate gave me the worst heartburn. Why, Obama, why. I'll never forgive you!
 

Valhelm

contribute something
I was working on a medium-effort post in the Gary Johnson thread, but it was closed. So I think I'll drop it here.

I find it's the same people with the "burn it all down" mentality. They don't want to vote the big parties and know that they are throwing their vote away.

Protest voting is harmful, but we shouldn't think of these folks as a bunch of edgy deplorables who want to cause disarray. Young Johnson and Stein voters, as well as many millennials who went for Bernie, are people who are aware that our society is burdened by a host of new and old problems, but are too poorly-informed to know how these problems should be solved.

A really common attitude is "all politicians are bad". This kind of simplistic thinking isn't limited to millennials, but you see it all the time on Gaf, on Reddit, or in the classroom. I used to work with a 24-year-old political teetotaler who sincerely thought that American politics were so corrupt that a vote would always be meaningless. Just being a politician was enough of a stain.

I think that because political awareness and historical understanding is so poor, most Americans -- particularly the young -- can't verbalize their frustrations, let alone how these problems have been solved. The main reason why I supported Bernie Sanders was not because I thought his policies would be revolutionary, but because he was a real step in the right direction. By starting the discussion on problems most candidates don't want to acknowledge, Bernie helped young Americans understand why they can't afford a degree, a wedding, or even their own apartment. He offered some real solutions to their problems, even if free college and single-payer healthcare might not be possible in our current climate.

The burden to help young people understand and ultimately solve brewing economic problems rests on the shoulders of millennial politicians. Just like Barry Goldwater fermented the far-right path that Reagan-era politicians would drag this country down, I think Bernie Sanders will leave a small, important mark on America's ideological history, and be remembered as the person who taught our generation that we don't need to rely on the whims of the private sector when hard times roll around.
 
But this isn't true! We have all sorts of reasons. For example, we have strong reason to suppose landline polls won't produce an accurate sample because there are many eligible voters who don't have a landline.



I might, actually. I think that's a pretty terrible conclusion he's made there. It's the statistician's equivalent of "eh, fuck it, just fudge it a bit and it should turn out okay".

Of course landline only is a bad methodology. Cohn's point was that within the band of reasonable assumptions, there is still a ton of variance. We know what's good and what's bad, but we don't know what's best. I would love to see a conversation between you and him about it. I'm sure I would learn something.
 
I love how every time Silver screams about a dead heat, the polling slaps him in the face.

Whenever he starts getting super aggressive it's usually because he's backed himself into a corner and unwilling to ease up. His worst punditry seems to come from being overly defensive and prideful.
 

Revolver

Member
I heard Jack Kingston on NewsHour claiming Hillary had 13 private email servers and it just went unchallenged by Judy Woodruff. Gah!
 

Joeytj

Banned
I heard Jack Kingston on NewsHour claiming Hillary had 13 private email servers and it just went unchallenged by Judy Woodruff. Gah!

Because she doesn't know for sure and she's used to campaigns making sense. Most anchors don't have sufficient facts ready to contradict so many lies, so they just stay quiet. The Trump campaign knows this.
 

Rebel Leader

THE POWER OF BUTTERSCOTCH BOTTOMS
I love that Hillary campaign has practiced in case they get insane Trump or subdued Trump. It makes my heart flutter
Subdued trump only shows up when teleprompters are there.

There is none
--

Truned on msnbs and saw
"Trump: moderator should not fact check me"

Seriously?
 
But this isn't true! We have all sorts of reasons. For example, we have strong reason to suppose landline polls won't produce an accurate sample because there are many eligible voters who don't have a landline.

This example is the equivalent of staking a moral claim in an argument on an extreme example. Arguments get harder with shades of grey.

In this case, it's less about methodologies that most would agree are bad, and more about controversial methods (i.e. methods that have a fair amount of both supporters and detractors). In these instances, instead of statisticians bickering over these technical issues, you average them together and do better than a coin flip.

In your previous example, for instance, where you had polls showing Trump +1, Clinton +1, Clinton +5, the assumption from you was that the Trump +1 poll was correct, so we shouldn't average these results, but take that poll. But what's your argument for doing so? It's obvious is the others were clearly bad (maybe landline only or something), but that isn't very interesting.

It's more accurate to average them and get Clinton +2 than it is to just guess which of the three is correct. A person guessing may pick the Clinton +5 poll, and that would be worse. (Obviously you aren't truly guessing but picking a method to stick with, but if it's not very clear which is best, you aren't going to have many readers interested in your vague defenses of the model. This is in fact my current frustration with Silver! I don't particularly know whether his model right now is correct or not, and won't until the election if ever, but his reasoning for why he does what he does is really poor.)

I stand by my statement before that Silver was probably great in calculation courses, but his arguments for things are lacking badly. Looking back, I was biased to this before when the results were in my favor (like, I was scared of Trump, so I didn't actually want him to win, so I liked Silver's model), but he was doing it then too (as he eventually admitted).
 

Wilsongt

Member
Time to start the long, painful task of download FFXI onto my new computer.

Polls swinging in Trump's favor? is the news media really so eager to bring about the end of the US just for some fucking ratings and website clicks?

Disgusting.,
 
Most Socialists are too eager to say "everything Trump has been accusing of doing wrt destroying U.S. norms, W already did!" but the media probably should look more at how post-truth politics came from Newt Gingrich and W post-9/11.
 
You know I just realized that usually even if you have democratic and republican pundits on, one side will usually at the very least admit their candidate didn't do as well of job as they liked on a certain issue or two. There is no way any of the Trump people will will be willing to move an inch even if Trump bombs catastrophically.
 
D

Deleted member 231381

Unconfirmed Member
You're right, FWIW. But...excluding the obvious, I think his point is that we don't know which voter turnout model is right and so we just have to average those (though I believe things like house effect should be included too).

But why should we average them? Do we have any statistical reason to suppose that averaging them gives us a more accurate result? No. We don't know which voter turnout model is right, I agree. But that doesn't mean we should therefore create an entirely new voter turnout model by averaging the all the other voter turnout models. Rather, we should either present the one we find most credible, or present multiple models having eliminated all those we find incredible. At least then, we can compare the different models to the final turnout, and find out whether our credibility was well-founded or not, rather than having one smushy model that makes examining whether individual assumptions worked or not very difficult.

Nate sort of does this. This is what the adjusted leader business is. He has examined how different methodologies fared in prior elections, tried to correct results produced by these methodologies accordingly to produce results that are tantamount to being produced by the same methodology, and then averaged those. But people have spent the last few pages attacking this. Baffling.
 

Sibylus

Banned
You know I just realized that usually even if you have democratic and republican pundits on, one side will usually at the very least admit their candidate didn't do as well of job as they liked on a certain issue or two. There is no way any of the Trump people will will be willing to move an inch even if Trump bombs catastrophically.

Everything will be the media's fault. Sad!
 
Well, I'll be posting here instead of the Debate thread. That thread will just be going way too fast for me.

Between twitter and my fantasy football scoring updates, gonna be too hard to manage two threads and so I'll stay here!

See y'all just before the debate!
 

Amir0x

Banned
Well, I'll be posting here instead of the Debate thread. That thread will just be going way too fast for me.

Between twitter and my fantasy football scoring updates, gonna be too hard to manage two threads and so I'll stay here!

See y'all just before the debate!
Don't worry we'll make sure this thread goes waaay too fast as well. Sorry Mamba :(
 
Status
Not open for further replies.
Top Bottom