Ornery.org
  Front Page   |   About Ornery.org   |   World Watch   |   Guest Essays   |   Contact Us

The Ornery American Forum Post New Topic  Post A Reply
my profile login | register | search | faq | forum home

  next oldest topic   next newest topic
» The Ornery American Forum » General Comments » States to watch (Page 2)

 - UBBFriend: Email this page to someone!   This topic comprises 4 pages: 1  2  3  4   
Author Topic: States to watch
JWatts
Member
Member # 6523

 - posted      Profile for JWatts   Email JWatts   Send New Private Message       Edit/Delete Post   Reply With Quote 
The race is pretty close to even. (Despite Nate Silver's hilariously partisan blogs masquerading as the mathematical Truth. Is Mr. Silver still claiming that Obama has a 70%+ chance of winning?).

This from Pew today:
quote:

As the presidential campaign enters its final week, Barack Obama has failed to regain much of the support he lost in the days following the first presidential debate and the race is now even among likely voters: 47% favor Obama while an identical percentage supports Mitt Romney.

So, in all likelihood the race will come down to Ohio. It's improbable that either candidate will win at this point without carrying Ohio.
Posts: 4700 | Registered: Oct 2009  |  IP: Logged | Report this post to a Moderator
DonaldD
Member
Member # 1052

 - posted      Profile for DonaldD   Email DonaldD   Send New Private Message       Edit/Delete Post   Reply With Quote 
74.6%
Posts: 10751 | Registered: May 2003  |  IP: Logged | Report this post to a Moderator
Wayward Son
Member
Member # 210

 - posted      Profile for Wayward Son   Email Wayward Son   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Despite Nate Silver's hilariously partisan blogs masquerading as the mathematical Truth. Is Mr. Silver still claiming that Obama has a 70%+ chance of winning?
74.6% chance for election day as of 10:57 PM on Oct. 28. (80.1% chance if the election were held today.) But, of course, that is "chance," as in a dice roll, and is based on the sum of all the polls, tweaked by his own formula to account for possible problems that he sees in the polls.

It ain't writ in stone, but it seems to be the best guess based on data and math rather than partisanship. After reading a few of the blogs, you can see that he is far more concerned with trying to be accurate than the particular outcome.

Do you have a better guess? If so, what is it based upon?

Posts: 8681 | Registered: Dec 2000  |  IP: Logged | Report this post to a Moderator
TCB
Member
Member # 1677

 - posted      Profile for TCB         Edit/Delete Post   Reply With Quote 
JWatts said:
quote:
The race is pretty close to even. (Despite Nate Silver's hilariously partisan blogs masquerading as the mathematical Truth. Is Mr. Silver still claiming that Obama has a 70%+ chance of winning?).
Just curious - do you think any prediction methodology that results in something other than an approximately 50% chance of an Obama victory is partisan? Or is it Silver's methodology in particular that you disagree with?
Posts: 824 | Registered: Apr 2004  |  IP: Logged | Report this post to a Moderator
G3
Member
Member # 6723

 - posted      Profile for G3   Email G3       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by TCB:
If Romney wins PA then he's already won every other swing state, plus a few other "leans Democrat" states. The end result would be 2008 in reverse. I doubt any Romney insiders have even dared to dream that big. [Smile]

I think it's more likely they're sitting on so much cash right now that they're not sure what to do with it all, and figured spending money in PA would generate favorable headlines that could help them in more competitive states.

Maybe, my guess is its more likely he's forcing Barry to spend here too now which siphons off from other states.
Posts: 2234 | Registered: May 2012  |  IP: Logged | Report this post to a Moderator
G3
Member
Member # 6723

 - posted      Profile for G3   Email G3       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by TCB:
JWatts said:
quote:
The race is pretty close to even. (Despite Nate Silver's hilariously partisan blogs masquerading as the mathematical Truth. Is Mr. Silver still claiming that Obama has a 70%+ chance of winning?).
Just curious - do you think any prediction methodology that results in something other than an approximately 50% chance of an Obama victory is partisan? Or is it Silver's methodology in particular that you disagree with?
Well, I think anything so out of whack with all polling data is highly suspect. Barry can't seem to crack 48% in Ohio,how the guy translates that to a greater than 70% chance is more about wishful thinking than actual math. All polling puts it in a dead heat with mitt slowly trending to advantage, silver seems to ignore that reality.
Posts: 2234 | Registered: May 2012  |  IP: Logged | Report this post to a Moderator
Wayward Son
Member
Member # 210

 - posted      Profile for Wayward Son   Email Wayward Son   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Barry can't seem to crack 48% in Ohio,how the guy translates that to a greater than 70% chance is more about wishful thinking than actual math. All polling puts it in a dead heat with mitt slowly trending to advantage, silver seems to ignore that reality.
You really should familiarize yourself with his methology before hurdling yourself to that conclusion. [Smile]
Posts: 8681 | Registered: Dec 2000  |  IP: Logged | Report this post to a Moderator
hobsen
Member
Member # 2923

 - posted      Profile for hobsen   Email hobsen   Send New Private Message       Edit/Delete Post   Reply With Quote 
Florida and Pennsylvania are close to opposites. Since Romney has a strong advantage in Florida, it would be unclear why he made a campaign appearance there recently - but his advisers no doubt fear that a major Obama effort might sway enough black and Spanish speaking voters to put the state in play before they realized they had trouble.

The opposite holds in Pennsylvania, which should be solid for Obama. But his advisers want to make sure the state stays theirs, so they are matching Republican efforts.

What both states have in common is that they are very expensive, and any candidate who wants to change how they lean would have to spend a lot of money and time. So any efforts in those states will probably be bluffs, or desperation moves. Such have some attraction because less attention and fewer polls mean one candidate or another might be better off than appears on the surface, so a major effort could just possibly pay off.

Posts: 4387 | Registered: Jul 2006  |  IP: Logged | Report this post to a Moderator
JWatts
Member
Member # 6523

 - posted      Profile for JWatts   Email JWatts   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by TCB:
JWatts said:
quote:
The race is pretty close to even. (Despite Nate Silver's hilariously partisan blogs masquerading as the mathematical Truth. Is Mr. Silver still claiming that Obama has a 70%+ chance of winning?).
Just curious - do you think any prediction methodology that results in something other than an approximately 50% chance of an Obama victory is partisan? Or is it Silver's methodology in particular that you disagree with?
Nate goes through a lot of statistical math and I have full confidence in his basic math abilities. But he openly admits that he weighs 'certain' polls heavier than others, using his own judgement. At that point, he's just choosing a set of assumptions and running with them. Garbage in, garbage out.

At this point in time he has a very short track record. In the election of 2008 he had direct access to the Obama inside polling information. So how much of the credit do you credit him and how much goes to the polling agency he was relying on? In the 2010 election he was ok, but certainly not exceptional in his predictions. It should be notable, that no one else is claiming anything like the confidence Nate's opinion columns evidence.

But the Left continues to quote his results as if he was an Oracle. It's confirmation bias big time. You can make the statistics say anything if you choose your own assumptions to fit a desired conclusion.

Posts: 4700 | Registered: Oct 2009  |  IP: Logged | Report this post to a Moderator
G3
Member
Member # 6723

 - posted      Profile for G3   Email G3       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by Wayward Son:
quote:
Barry can't seem to crack 48% in Ohio,how the guy translates that to a greater than 70% chance is more about wishful thinking than actual math. All polling puts it in a dead heat with mitt slowly trending to advantage, silver seems to ignore that reality.
You really should familiarize yourself with his methology before hurdling yourself to that conclusion. [Smile]
I have, and I have also familiarized myself with the other polls. Silver is on a rather massive outlier and as such is highly suspect. Maybe he got it right and every single other polling organization with decades of experience got it wrong this time but which seems the more likely?
Posts: 2234 | Registered: May 2012  |  IP: Logged | Report this post to a Moderator
G3
Member
Member # 6723

 - posted      Profile for G3   Email G3       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by hobsen:
The opposite holds in Pennsylvania, which should be solid for Obama. But his advisers want to make sure the state stays theirs, so they are matching Republican efforts.

That they even have to make sure it stays theirs indicates a rather massive shift mat be occurring. PA should have been as much a lock as NY and CA.

[ October 30, 2012, 10:04 AM: Message edited by: G3 ]

Posts: 2234 | Registered: May 2012  |  IP: Logged | Report this post to a Moderator
TomDavidson
Member
Member # 99

 - posted      Profile for TomDavidson   Email TomDavidson   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
PA should have been as much a lock as NY and CA.
*laugh*
Posts: 22935 | Registered: Nov 2000  |  IP: Logged | Report this post to a Moderator
TCB
Member
Member # 1677

 - posted      Profile for TCB         Edit/Delete Post   Reply With Quote 
JWatts said:
quote:
At this point in time he has a very short track record. In the election of 2008 he had direct access to the Obama inside polling information. So how much of the credit do you credit him and how much goes to the polling agency he was relying on? In the 2010 election he was ok, but certainly not exceptional in his predictions. It should be notable, that no one else is claiming anything like the confidence Nate's opinion columns evidence.
I see where you're coming from, but I think you're overstating Silver's confidence a bit - his forecast is probabilistic, not really predictive, and it still gives Romney a significant chance of winning. Also, he's not the only one who sees Obama as a moderate favorite - his odds are pretty close to those of betting markets, which, absent tampering, can't be anything except rational.

The analogy he gave a couple days ago was that the Giants lead the Redskins 24-21 late in the fourth quartee. Anyone could still win, but you can't really say it's a toss-up, either - the Giants are the favorites. We could trade a few posts convoluting the analogy ("but the Redskins are in the red zone and Eli Manning is out with a concussion" and so on), but you get my point. [Smile]

Anyway, I know most conservatives think it's biased, and it's possible that it is. But Silver actually does have interesting insights. No one else (that I read, at least) is writing about the election from the same point of view.

Posts: 824 | Registered: Apr 2004  |  IP: Logged | Report this post to a Moderator
AI Wessex
Member
Member # 6653

 - posted      Profile for AI Wessex   Email AI Wessex   Send New Private Message       Edit/Delete Post   Reply With Quote 
The storm (aka Wetzilla) will have a psychological effect on the people of Virginia. Some analysts say those voters will viscerally blame Obama for their misfortune, others say they will look for stability in government as represented by him to help them overcome a natural catastrophe.

One thing they should know is that Romney has promised to cut FEMA spending and delegate the responsibility to the states for their own cleanup or privatize its operations altogether. Right now, the government will cover 75% of all costs associated with a declared disaster. What does Romney propose to do in its place?

Posts: 8393 | Registered: Feb 2011  |  IP: Logged | Report this post to a Moderator
Wayward Son
Member
Member # 210

 - posted      Profile for Wayward Son   Email Wayward Son   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Maybe he got it right and every single other polling organization with decades of experience got it wrong this time but which seems the more likely?
I don't understand who you are referring to. Nate uses the data from just about every single polling organization (Rasmussen, PPP, NBC, Fox, CNN, Grove, etc., etc.) and synthesizes it. He does not do his own polling.

Which polling organizations do you trust more than Nate and why?

quote:
But he openly admits that he weighs 'certain' polls heavier than others, using his own judgement. At that point, he's just choosing a set of assumptions and running with them. Garbage in, garbage out.
As Nate's columns discuss, you have to use your own judgement when looking at the polls. Some of them tend to lean toward one side or another (if only the candidate's own polls [Smile] ), so you have to give some more or less credibility than others. You'll tend to be wrong more times without judgement than with it.

quote:
It should be notable, that no one else is claiming anything like the confidence Nate's opinion columns evidence.
[shug] It may be confidence based on hard work, knowledge and data. It may be arrogance. It may be that, even with his confidence, he is predicting that Romney would win a quarter of the time.

What impresses me is not his confidence or the results he predicts. What impresses me is his detailed column of why he judges things as he does. He talks about his methodology, his judgements, his opinions, and how they influence his conclusions. He talks about likely-voter vs. all voter surveys and how much confidence to give in each. He talks about the influence of economics, of race, of local politics, and how they add up in his estimation. He tells you about all the factors he is considering.

Who else does so in that detail? Who else puts their thinking right out there, so you can judge yourself?

When someone is going into that much detail in his analysis, partisan bias tends to get lost. It's hard to keep things leaning on one side or the other when you are trying to weigh how much confidence you should give state polls vs. national polls, especially compared to off-the-cuff estimates that so many other commentators do. His focus on the math and statistics makes bias much less likely.

quote:
But the Left continues to quote his results as if he was an Oracle. It's confirmation bias big time. You can make the statistics say anything if you choose your own assumptions to fit a desired conclusion.
I'm sure the Left would be much less likely to quote him if he were predicting a Romney victory.

But then I bet the Right would have a whole lot more confidence in him, too. [Smile]

There is definitely a danger of confirmational bias with him. But there is that danger with anyone. As I stated before, Nate's obsession with statistics and details, and his daily essays about his methodology, means that there is probably less confirmational bias than most commentators. If you know of someone better, let me know. I would love to read his opinions, too.

The News Media likes to talk to him because he is a wealth of information. They can get him to explain his reasoning for as long as they need him to. Makes good TV.

Don't discount him just for that. He isn't perfect, but no one is. He just seems to be the best out there right now. [Smile]

Posts: 8681 | Registered: Dec 2000  |  IP: Logged | Report this post to a Moderator
JoshuaD
Member
Member # 1420

 - posted      Profile for JoshuaD   Email JoshuaD   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by TCB:
If Romney wins PA then he's already won every other swing state, plus a few other "leans Democrat" states. The end result would be 2008 in reverse. I doubt any Romney insiders have even dared to dream that big. [Smile]

I think it's more likely they're sitting on so much cash right now that they're not sure what to do with it all, and figured spending money in PA would generate favorable headlines that could help them in more competitive states.

PA is Mr. Romney's back up plan. Ohio, Iowa, Wisconsin, Nevada, and Colorado have been worked over by both candidates so much that there's not much more ground to gain there. So he's looking for a few other, higher ROI, inroads. I don't think anyone expects PA to turn towards Mr. Romney, but I think the analysis is "If any state is loose, let's find it." There's a small chance PA is, so they're taking a stab.
Posts: 3742 | Registered: Dec 2003  |  IP: Logged | Report this post to a Moderator
JoshuaD
Member
Member # 1420

 - posted      Profile for JoshuaD   Email JoshuaD   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by G3:
quote:
Originally posted by TCB:
JWatts said:
quote:
The race is pretty close to even. (Despite Nate Silver's hilariously partisan blogs masquerading as the mathematical Truth. Is Mr. Silver still claiming that Obama has a 70%+ chance of winning?).
Just curious - do you think any prediction methodology that results in something other than an approximately 50% chance of an Obama victory is partisan? Or is it Silver's methodology in particular that you disagree with?
Well, I think anything so out of whack with all polling data is highly suspect. Barry can't seem to crack 48% in Ohio,how the guy translates that to a greater than 70% chance is more about wishful thinking than actual math. All polling puts it in a dead heat with mitt slowly trending to advantage, silver seems to ignore that reality.
His methodology seems really solid. It's a very close race, but it's also a very well-understood race. There's not a ton of room for variance in the polls. So even a small edge spread out over a large number of polls can result in a significantly high chance of winning. Appearing to have 49% of the popular vote doesn't translate to having a 49% chance of winning the election. On election day, 49% of the vote gets you 0% of the presidency.
Posts: 3742 | Registered: Dec 2003  |  IP: Logged | Report this post to a Moderator
JoshuaD
Member
Member # 1420

 - posted      Profile for JoshuaD   Email JoshuaD   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Which polling organizations do you trust more than Nate and why?
I tend to think Nate's a little bit optimistic for the President, for what it's worth. It's a difficult thing to be completely non-biased, and I see a small bias (in desired outcome) sneak into the tone of his articles here and there. I tend to think that this is probably reflected a little in his analysis of the polls. Giving a certain set of polls just a little more weight, or another a little less weight, could really change the analysis of his algorithm.

That being said, he's the best out there, IMO.

[ October 30, 2012, 12:17 PM: Message edited by: JoshuaD ]

Posts: 3742 | Registered: Dec 2003  |  IP: Logged | Report this post to a Moderator
JoshuaD
Member
Member # 1420

 - posted      Profile for JoshuaD   Email JoshuaD   Send New Private Message       Edit/Delete Post   Reply With Quote 
I'll be very curious how this storm will affect the election in Virginia. There's not much for the President to lose there, but he can win everything if he picks up 1 voter out of 50.
Posts: 3742 | Registered: Dec 2003  |  IP: Logged | Report this post to a Moderator
JWatts
Member
Member # 6523

 - posted      Profile for JWatts   Email JWatts   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by JoshuaD:
I'll be very curious how this storm will affect the election in Virginia. There's not much for the President to lose there, but he can win everything if he picks up 1 voter out of 50.

Sure, if Obama could pick up an extra 2% he certainly could win the election, though not everything. But at this point even NPR has the President down by 1%. If the Left has started hoping that a hurricane will upset Romney's momentum, it's a sign of desperation.

Romney has started Ad spending in PA, Mason Dixon has stopped polling in FL (Romney has it locked in their opinion) and at this point the election is pretty much down to Ohio.

As to Nate Silver, I've got a simple check for bias. His entire model is based upon estimating the percentage vote per state, multiplying by the electoral vote and calculating the expected chance of winning. So if his model isn't biased, than after the election, regardless of who wins, his model should show an equal amount of over/under results for each state. Ergo, his state predictions should have 25 states voting over his predicted number for Obama and 25 states under his prediction for Obama. If there is no bias. If his model over predicts the Obama vote in more than 25 states, he's biased, not accurate. Simple enough to double check after the election.

Posts: 4700 | Registered: Oct 2009  |  IP: Logged | Report this post to a Moderator
yossarian22c
Member
Member # 1779

 - posted      Profile for yossarian22c   Email yossarian22c       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by JWatts:
As to Nate Silver, I've got a simple check for bias. His entire model is based upon estimating the percentage vote per state, multiplying by the electoral vote and calculating the expected chance of winning. So if his model isn't biased, than after the election, regardless of who wins, his model should show an equal amount of over/under results for each state. Ergo, his state predictions should have 25 states voting over his predicted number for Obama and 25 states under his prediction for Obama. If there is no bias. If his model over predicts the Obama vote in more than 25 states, he's biased, not accurate. Simple enough to double check after the election.

Actually I'll nitpick a little here. There is not the same amount of information for all 50 states. If his model has a small and balanced margin of error for the swing states then I would say his methodology is good. If the predictions are much more skewed in Alabama, Mississippi and California then it is likely the lack of information that caused the error, not some inherent bias in the method.
Posts: 1121 | Registered: May 2004  |  IP: Logged | Report this post to a Moderator
DonaldD
Member
Member # 1052

 - posted      Profile for DonaldD   Email DonaldD   Send New Private Message       Edit/Delete Post   Reply With Quote 
I'm pretty sure Joshua wouldn't characterize himself as being part of "the left" [Smile]
Posts: 10751 | Registered: May 2003  |  IP: Logged | Report this post to a Moderator
JWatts
Member
Member # 6523

 - posted      Profile for JWatts   Email JWatts   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by yossarian22c:
Actually I'll nitpick a little here. There is not the same amount of information for all 50 states. If his model has a small and balanced margin of error for the swing states then I would say his methodology is good. If the predictions are much more skewed in Alabama, Mississippi and California then it is likely the lack of information that caused the error, not some inherent bias in the method.

That shouldn't cause his polling results to skew towards Obama though (or Romney either). I'm not saying measure his range of error, just check if he was over/under. It's a 50/50 proposition. If there isn't any bias, the results should be even. If not, then he's got a bad model.
Posts: 4700 | Registered: Oct 2009  |  IP: Logged | Report this post to a Moderator
PSRT
Member
Member # 6454

 - posted      Profile for PSRT   Email PSRT   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
So if his model isn't biased, than after the election, regardless of who wins, his model should show an equal amount of over/under results for each state.
One could also check his work from 2008, and one would find that in that election, his results under-reported how well Obama would do, suggesting that even if he is biased, its not showing up in his models, and that claims of a pro-Obama bias are more wishful thinking than data driven.
Posts: 2152 | Registered: Apr 2009  |  IP: Logged | Report this post to a Moderator
JoshCrow
Member
Member # 6048

 - posted      Profile for JoshCrow   Email JoshCrow   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by G3:
I have, and I have also familiarized myself with the other polls. Silver is on a rather massive outlier and as such is highly suspect. Maybe he got it right and every single other polling organization with decades of experience got it wrong this time but which seems the more likely?

Oh but only if you applied this kind of thinking to climate change...
Posts: 2281 | Registered: Mar 2008  |  IP: Logged | Report this post to a Moderator
G3
Member
Member # 6723

 - posted      Profile for G3   Email G3       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by JoshCrow:
quote:
Originally posted by G3:
I have, and I have also familiarized myself with the other polls. Silver is on a rather massive outlier and as such is highly suspect. Maybe he got it right and every single other polling organization with decades of experience got it wrong this time but which seems the more likely?

Oh but only if you applied this kind of thinking to climate change...
The is no relation between the two. AGW has been exposed as a fraud via a whistle-blower and multiple scientific reviews. Silver's methodology may provide misleading results in this case but, as far as I know, it's not fraudulent.
Posts: 2234 | Registered: May 2012  |  IP: Logged | Report this post to a Moderator
PSRT
Member
Member # 6454

 - posted      Profile for PSRT   Email PSRT   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
AGW has been exposed as a fraud via a whistle-blower and multiple scientific reviews
Only to ridiculously biased observers, who I wouldn't trust to know bias in others if it hit them in the face with a trout.
Posts: 2152 | Registered: Apr 2009  |  IP: Logged | Report this post to a Moderator
JoshCrow
Member
Member # 6048

 - posted      Profile for JoshCrow   Email JoshCrow   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by G3:
quote:
Originally posted by JoshCrow:
quote:
Originally posted by G3:
I have, and I have also familiarized myself with the other polls. Silver is on a rather massive outlier and as such is highly suspect. Maybe he got it right and every single other polling organization with decades of experience got it wrong this time but which seems the more likely?

Oh but only if you applied this kind of thinking to climate change...
The is no relation between the two. AGW has been exposed as a fraud via a whistle-blower and multiple scientific reviews. Silver's methodology may provide misleading results in this case but, as far as I know, it's not fraudulent.
Nevertheless, if your position is that the majority of climatologists with "decades of experience" are conducting a fraud, why would you suddenly give the benefit of the doubt to a majority of pollsters (who, it seems to me, have also given Obama the edge in odds anyhow)?

You seem pretty quick to lend credence to a "majority" view when it suits you, or oppose it when it does not.

Posts: 2281 | Registered: Mar 2008  |  IP: Logged | Report this post to a Moderator
Wayward Son
Member
Member # 210

 - posted      Profile for Wayward Son   Email Wayward Son   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
I'm not saying measure his range of error, just check if he was over/under. It's a 50/50 proposition. If there isn't any bias, the results should be even. If not, then he's got a bad model.
Certainly, if it is within reason. After all, flipping a coin 50 times won't get you 25 heads and 25 tails most of the time. [Wink]
Posts: 8681 | Registered: Dec 2000  |  IP: Logged | Report this post to a Moderator
JWatts
Member
Member # 6523

 - posted      Profile for JWatts   Email JWatts   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by PSRT:
quote:
So if his model isn't biased, than after the election, regardless of who wins, his model should show an equal amount of over/under results for each state.
One could also check his work from 2008, and one would find that in that election, his results under-reported how well Obama would do, suggesting that even if he is biased, its not showing up in his models, and that claims of a pro-Obama bias are more wishful thinking than data driven.
So, $50 bet (to charity) if Mr. Silver shows no pro-Obama bias in his current listed predictions vs the final election tally then?
Posts: 4700 | Registered: Oct 2009  |  IP: Logged | Report this post to a Moderator
Pyrtolin
Member
Member # 2638

 - posted      Profile for Pyrtolin   Email Pyrtolin   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by JWatts:
quote:
Originally posted by yossarian22c:
Actually I'll nitpick a little here. There is not the same amount of information for all 50 states. If his model has a small and balanced margin of error for the swing states then I would say his methodology is good. If the predictions are much more skewed in Alabama, Mississippi and California then it is likely the lack of information that caused the error, not some inherent bias in the method.

That shouldn't cause his polling results to skew towards Obama though (or Romney either). I'm not saying measure his range of error, just check if he was over/under. It's a 50/50 proposition. If there isn't any bias, the results should be even. If not, then he's got a bad model.
It doesn't have to be perfect in any given case- the test of how useful the formula is would be in whether it, on average, is less biased than the polling data that it draws from.
Posts: 11997 | Registered: Oct 2005  |  IP: Logged | Report this post to a Moderator
G3
Member
Member # 6723

 - posted      Profile for G3   Email G3       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by JoshCrow:
quote:
Originally posted by G3:
quote:
Originally posted by JoshCrow:
quote:
Originally posted by G3:
I have, and I have also familiarized myself with the other polls. Silver is on a rather massive outlier and as such is highly suspect. Maybe he got it right and every single other polling organization with decades of experience got it wrong this time but which seems the more likely?

Oh but only if you applied this kind of thinking to climate change...
The is no relation between the two. AGW has been exposed as a fraud via a whistle-blower and multiple scientific reviews. Silver's methodology may provide misleading results in this case but, as far as I know, it's not fraudulent.
Nevertheless, if your position is that the majority of climatologists with "decades of experience" are conducting a fraud, why would you suddenly give the benefit of the doubt to a majority of pollsters (who, it seems to me, have also given Obama the edge in odds anyhow)?

You seem pretty quick to lend credence to a "majority" view when it suits you, or oppose it when it does not.

Do you think our understanding of polling is the same as our understanding of planetary atmospherics? You're making comparisons that are absurd.
Posts: 2234 | Registered: May 2012  |  IP: Logged | Report this post to a Moderator
TCB
Member
Member # 1677

 - posted      Profile for TCB         Edit/Delete Post   Reply With Quote 
JWatts said:
quote:
As to Nate Silver, I've got a simple check for bias. His entire model is based upon estimating the percentage vote per state, multiplying by the electoral vote and calculating the expected chance of winning. So if his model isn't biased, than after the election, regardless of who wins, his model should show an equal amount of over/under results for each state.
The issue I see with that is that the 538 model is essentially a poll aggregator - like the RealClearPolitics average, but fancier - with some economic indicators baked in. If the combination of state and national polls are systematically biased, Silver's model will reflect the same bias (garbage in, garbage out).

But if the polls are an accurate reflection of the election results, then, your're right - if his model is good it won't show any systematic bias.

Given all the commentary Silver has attracted in the past week, I expect we'll see plenty of post-mortems late next week, either from his defenders or detractors, depending on how the voting goes. [Smile]

Posts: 824 | Registered: Apr 2004  |  IP: Logged | Report this post to a Moderator
Pyrtolin
Member
Member # 2638

 - posted      Profile for Pyrtolin   Email Pyrtolin   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by TCB:
The issue I see with that is that the 538 model is essentially a poll aggregator - like the RealClearPolitics average, but fancier - with some economic indicators baked in. If the combination of state and national polls are systematically biased, Silver's model will reflect the same bias (garbage in, garbage out).

The active issue at hand is, actually, whether his attempts to identify and correct for such bias thorugh weighting and other adjustments are evenhanded or whether they actively introduce his own bias to the process. If his system is good, it should be able to take garbage and turn it into something useful. If his system is bad then it will have its own bias that has to be corrected for (if not making it outright inaccurate and thus completely unreliable)
Posts: 11997 | Registered: Oct 2005  |  IP: Logged | Report this post to a Moderator
AI Wessex
Member
Member # 6653

 - posted      Profile for AI Wessex   Email AI Wessex   Send New Private Message       Edit/Delete Post   Reply With Quote 
Isn't it pretty simple really? The RCP poll average vs the 538 analysis based on the polls. It should be pretty simple to see which model was more predictive.

FWIW, my opinion is that polls have value only in terms of their distance from the election itself. Thus, a poll a month ago represents a two month "trend", one 2 months ago represents 4 months. It doesn't matter if one week in the middle of those time spans is different from the others, as the average over the entire period matters. Now that we're within a week, a daily fluctuation might not matter, but 4 or even 3 days of trending is pretty significant.

If that's so, then Romney's lead in the popular vote is within the margin of error, but Obama's lead in the electoral college is not. It will be *very* hard for Mitt to win, no matter how the popular vote goes.

One last FWIW, I am one of those rare people who actually likes the electoral college system. If we all lived in a single representative environment, then the popular vote would be preferable. But - in theory - each state has a guaranteed representation in our government and in the national election. I've never even been to Wyoming, but I'm very happy that they have their electors!

Posts: 8393 | Registered: Feb 2011  |  IP: Logged | Report this post to a Moderator
G3
Member
Member # 6723

 - posted      Profile for G3   Email G3       Edit/Delete Post   Reply With Quote 
More polls! Don't you just love'em! And a little analysis, from this morning:
quote:
Let’s look at the Ohio internals. Obama up by 5 with a D+8 turnout. Hmm. It was D+5 in 2008. Is D+8 likely this time around? Probably not. Also, the same poll has Romney winning Independents by 6 points, which is about the same margin that Obama won Independents by in 2008. Yet we are supposed to believe he’ll do better this time overall with a 13 point swing in the Independent vote?
and

quote:
The same polling company (CBS/NYT/Quinnipiac) released a poll in Virginia too. Obama apparently leads by 2 points. Yet at the same time he is losing the Independent vote by 21 points! The split? D+8.
Mitt's killing it in the independent vote but Barry holds the overall lead? Heavy, and potentially unsupportable, advantages to Barry in the sample and Barry is still just barely holding onto enough to call it a tossup. I think it's either much closer than the poll says or Mitt has the edge.
Posts: 2234 | Registered: May 2012  |  IP: Logged | Report this post to a Moderator
Wayward Son
Member
Member # 210

 - posted      Profile for Wayward Son   Email Wayward Son   Send New Private Message       Edit/Delete Post   Reply With Quote 
Oddly enough, Nate discusses the current discrepancy between the state polls and the national polls in today's blog. He even gives a table of the state polling estimates from his model and six other state-centric models.

And he agrees with you, JWatts.

quote:
Bias, in a statistical sense, means missing consistently in one direction — for example, overrating the Republican’s performance across a number of different examples, or the Democrat’s...

In a previous article, I examined the history of bias in public polls based on whether they’ve tended systematically to overrate the standing of the Democrat or the Republican. (The answer is that they don’t exhibit either bias on a consistent basis, as long as you’re using likely voter polls; registered voter polls will tend to overstate the vote for the Democrat.)

quote:
Given all the commentary Silver has attracted in the past week, I expect we'll see plenty of post-mortems late next week, either from his defenders or detractors, depending on how the voting goes.
And for the most detailed and comprehensive post-mortem, check out Nate's site. He'll dissect his model the most. [Smile]
Posts: 8681 | Registered: Dec 2000  |  IP: Logged | Report this post to a Moderator
JoshuaD
Member
Member # 1420

 - posted      Profile for JoshuaD   Email JoshuaD   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by Pyrtolin:
If his system is good, it should be able to take garbage and turn it into something useful. If his system is bad then it will have its own bias that has to be corrected for (if not making it outright inaccurate and thus completely unreliable)

It doesn't work like that. You can't recover good information from garbage. The best yo can do is make guesses to identify the bad info and fill in the gaps, but you can't tease reliable data out from unreliable data in the way you're suggesting.
Posts: 3742 | Registered: Dec 2003  |  IP: Logged | Report this post to a Moderator
TCB
Member
Member # 1677

 - posted      Profile for TCB         Edit/Delete Post   Reply With Quote 
Pyrtolin said:
quote:
The active issue at hand is, actually, whether his attempts to identify and correct for such bias thorugh weighting and other adjustments are evenhanded or whether they actively introduce his own bias to the process. If his system is good, it should be able to take garbage and turn it into something useful. If his system is bad then it will have its own bias that has to be corrected for (if not making it outright inaccurate and thus completely unreliable)
I think that's true for particular polls. He's adjusting for polls not including cellphone-only voters, for instance, or a poll not screening for likely voters, or a poll having a partisan lean relative to the consensus. All that works for correcting an individual poll.

But Silver is assuming that the polls are, taken together, accurate (with the addition of a bit of economic data). If the collective polls are systematically biased in some way, then Silver's analysis will reflect the same skew. For example, if G3 is right and the electorate is very close to 2010's then Romney will win a landslide, and Silver's forecast will just as ridiculous as the polls are.

Posts: 824 | Registered: Apr 2004  |  IP: Logged | Report this post to a Moderator
Pyrtolin
Member
Member # 2638

 - posted      Profile for Pyrtolin   Email Pyrtolin   Send New Private Message       Edit/Delete Post   Reply With Quote 
quote:
Originally posted by JoshuaD:
quote:
Originally posted by Pyrtolin:
If his system is good, it should be able to take garbage and turn it into something useful. If his system is bad then it will have its own bias that has to be corrected for (if not making it outright inaccurate and thus completely unreliable)

It doesn't work like that. You can't recover good information from garbage. The best yo can do is make guesses to identify the bad info and fill in the gaps, but you can't tease reliable data out from unreliable data in the way you're suggesting.
That would be true if we were talking about a single source of data- but we're talking about multiple sources of data with different levels of bias and reliability. If a particular poll happens to be unreliable, (I take it you mean "inaccurate" by that) then you weight it less, or, eventually exclude it completely if becomes clear that it's so inaccurate as to be useless. If a poll is consistently biased (like, say, Rasmussen's consistent 3 point Republican bias) then you can apply a correction factor.

The entire point of aggregating them is explicitly so that you can fill in those gaps, and the correction factors that you apply on top of that are explicitly part of the process of identifying and filling the gaps.

Posts: 11997 | Registered: Oct 2005  |  IP: Logged | Report this post to a Moderator
  This topic comprises 4 pages: 1  2  3  4   

Quick Reply
Message:

HTML is not enabled.
UBB Code™ is enabled.
UBB Code™ Images not permitted.
Instant Graemlins
   


Post New Topic  Post A Reply Close Topic   Feature Topic   Move Topic   Delete Topic next oldest topic   next newest topic
 - Printer-friendly view of this topic
Hop To:


Contact Us | Ornery.org Front Page

Powered by Infopop Corporation
UBB.classic™ 6.7.1