If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Denver Post)   Election model that predicted the presidency with 100% accuracy for the past 30 years gets reset back to zero   (blogs.denverpost.com) divider line 112
    More: Followup, poll average, economic malaise, personal incomes, climate change denial, presidents, OpenID, electoral colleges, exit polls  
•       •       •

7009 clicks; posted to Politics » on 13 Nov 2012 at 2:50 PM (1 year ago)   |  Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



112 Comments   (+0 »)
   
View Voting Results: Smartest and Funniest

Archived thread

First | « | 1 | 2 | 3 | » | Last | Show all
 
2012-11-13 01:11:08 PM
Its not %100 accurate when the model didn't exist for those thirty years.
 
2012-11-13 01:13:22 PM
"The model says there's a correlation between economic conditions and the presidential vote," Bickers said. "What it means is the president in a sense rose above gravity. He managed to do something that historically doesn't happen - an incumbent candidate not being as tied to the economic conditions as is normally the case."

I have an election forecast model that predicts in times of economic stress, candidates that act like Thurston Howell don't get elected.

It's 100% correct so far.
 
2012-11-13 01:19:02 PM

Lando Lincoln: "The model says there's a correlation between economic conditions and the presidential vote," Bickers said. "What it means is the president in a sense rose above gravity. He managed to do something that historically doesn't happen - an incumbent candidate not being as tied to the economic conditions as is normally the case."

I have an election forecast model that predicts in times of economic stress, candidates that act like Thurston Howell don't get elected.

It's 100% correct so far.


And running a campaign of "I'm not that guy" doesn't work. The left learned that in 2004. It remains to be seen if the GOP can learn from 2012. You don't get turnout voting against someone, you get turnout by voting FOR someone.
 
2012-11-13 01:25:12 PM
FTAThe tool ended up underestimating Obama by 5 to 6 percentage points

Is this his methodology or him?
 
2012-11-13 01:49:07 PM

Aarontology: Its not %100 accurate when the model didn't exist for those thirty years.


They "retroactively" ran it on those numbers.

I don't know how much old poll data you can find going back to the 80s, 70s, etc, (you can probably glean some data from old newspaper articles), but, would be interesting to see how Nate's formulas would have tracked past elections, and how accurate it would be.
 
2012-11-13 01:51:37 PM
Does the model take rape into account?
 
2012-11-13 02:01:14 PM
The model erred, Bickers said, in assuming that a campaign could not shake off such challenging economic factors. One thing lacking, he said, is a good way to model campaign impact.

No, the model erred because you have to go back at least to the 30s to find a similar economy. To think today's economic conditions could predict the election based off the economic conditions in last 7 election was foolish.

You might as well have cherry-picked the results of sporting events to make them fit back to the 80s. It would have been just as predictive.
 
2012-11-13 02:04:32 PM
Bickers said the Obama campaign managed to neutralize Romney's "strengths on economic stewardship" - exit polls showed voters held similar views on each candidate's ability to steer the economy - in part by shifting attention to issues such as immigration and women's reproductive rights that play to Obama's strengths.

Let's rewrite that.

Bickers said the Obama campaign managed to neutralize Romney's "strengths on economic stewardship" in part by shifting attention to issues such as immigration and women's reproductive rights that play to Obama's strengths. Exit polls showed voters held similar views on each candidate's ability to steer the economy.

Obama didn't neutralize Romney's "strength on the economy" with immigration and woman's issues, he neutralized the very idea that Romney would be strong for the economy.
 
2012-11-13 02:04:44 PM
"What it means is the president in a sense rose above gravity. He managed to do something that historically doesn't happen"

Obviously Bickers is not a Bayesian.
 
2012-11-13 02:32:44 PM

impaler: The model erred, Bickers said, in assuming that a campaign could not shake off such challenging economic factors. One thing lacking, he said, is a good way to model campaign impact.

No, the model erred because you have to go back at least to the 30s to find a similar economy. To think today's economic conditions could predict the election based off the economic conditions in last 7 election was foolish.

You might as well have cherry-picked the results of sporting events to make them fit back to the 80s. It would have been just as predictive.


The hell you say?
 
2012-11-13 02:42:10 PM

sweetmelissa31: Obviously Bickers is not a Bayesian.


LOL.

FTFA: "University of Colorado political science professor"

Well there's his problem.


Dear poli-sci department,

Leave predictive modeling to the statistics department.

Love,

Reality.
 
2012-11-13 02:46:57 PM
It's not just the economy, stupid professors.
 
2012-11-13 02:55:14 PM
Their model only proved two things:

1. They don't understand the meaning of the words "prediction" or "20/20 hindsight"

2. They missed the first day of Statistics 101 when the first words out of their professor's mouth were "correlation does not prove causation"
 
2012-11-13 02:55:24 PM

sweetmelissa31: "What it means is the president in a sense rose above gravity. He managed to do something that historically doesn't happen"

Obviously Bickers is not a Bayesian.


I wish I got math jokes :(
 
2012-11-13 02:58:11 PM
One of the problems of the internet is that we can collectively remember more than 72 hours of history. So when you say "this model accurately predicted X in the past" we can look and see if it's both true and actually used in the past rather than being post hoc data processing.
 
2012-11-13 02:58:33 PM
Based on the article alone I'd guess that guy is republican every quote of his was a softer more modest version of republican hyperbole.
 
2012-11-13 02:59:14 PM

GAT_00: And running a campaign of "I'm not that guy" doesn't work. The left learned that in 2004. It remains to be seen if the GOP can learn from 2012. You don't get turnout voting against someone, you get turnout by voting FOR someone.


Huh? That's the exact campaign Obama ran against Romney. "That guy's a rich, White, out-of-touch, lawbreaking killer."
 
2012-11-13 02:59:20 PM
His Neural Nets need restrung
 
2012-11-13 03:00:02 PM

GAT_00: Lando Lincoln: "The model says there's a correlation between economic conditions and the presidential vote," Bickers said. "What it means is the president in a sense rose above gravity. He managed to do something that historically doesn't happen - an incumbent candidate not being as tied to the economic conditions as is normally the case."

I have an election forecast model that predicts in times of economic stress, candidates that act like Thurston Howell don't get elected.

It's 100% correct so far.

And running a campaign of "I'm not that guy" doesn't work. The left learned that in 2004. It remains to be seen if the GOP can learn from 2012. You don't get turnout voting against someone, you get turnout by voting FOR someone.


they probably knew this, but no one could figure out how to talk people into voting FOR romney.
 
2012-11-13 03:00:14 PM
"I still think it's a useful tool. It didn't get the correct outcome but it does give you the historical context."

I have this thing that predicts things. I mean sure it predicts them wrong but I still find it useful.

"The model was wrong," Bickers conceded.

Yes, also you were wrong. This means you have zero credibility.
 
2012-11-13 03:02:32 PM
25.media.tumblr.com
 
2012-11-13 03:02:36 PM
It didn't predict jackshiat. It got one prediction wrong and correctly modeled 7 other presidential elections. It seems that the 8th election was nothing like the previous seven.

Nate Silver and that guy from Princeton have predicted two elections damn near exactly.

Making up a model and then tweaking it enough to make it fit already known data is not prediction.
 
2012-11-13 03:02:52 PM
This was not a prediction proved correct over the past 30 years. It was drawing a trend line and telling rubes that the next data point had to fall on the same line as the others.

If you think it's just the economy, you're stupid.
 
2012-11-13 03:02:57 PM
Nate Silver (of course) already addressed this.

There was a Colorado University model that correctly predicted the past 30 years of elections. However, two conservative leaning professors decided in 2012 to redo the model from scratch, and shockingly, it predicted a Romney win. But that particular model had never been used to predict anything before, so it had no streak to lose.
 
2012-11-13 03:03:08 PM
didn't the redskins one bite the dust this year too?
 
2012-11-13 03:03:11 PM

lennavan: "I still think it's a useful tool. It didn't get the correct outcome but it does give you the historical context."

I have this thing that predicts things. I mean sure it predicts them wrong but I still find it useful.

"The model was wrong," Bickers conceded.

Yes, also you were wrong. This means you have zero credibility.


If he continues his streak of being wrong and never breaks it he will regain credibility as a predictor of elections.
 
2012-11-13 03:03:14 PM

lennavan: I have this thing that predicts things. I mean sure it predicts them wrong but I still find it useful.


It predicts the past with 100% accuracy. You can't explain that.
 
2012-11-13 03:03:19 PM

jjorsett: GAT_00: And running a campaign of "I'm not that guy" doesn't work. The left learned that in 2004. It remains to be seen if the GOP can learn from 2012. You don't get turnout voting against someone, you get turnout by voting FOR someone.

Huh? That's the exact campaign Obama ran against Romney. "That guy's a rich, White, out-of-touch, lawbreaking killer."


That's odd. I searched Obama's Convention speech, you know, the one where he had a chance to speak to tens of millions of people across the nation and the word Romney didn't appear in it at all. Granted, I found it on NPR which we all know is on Obama's payroll, so they probably edited it.

Keep on fightin the good fight man. Hannity would be proud.
 
2012-11-13 03:03:25 PM
My back-tested model says all future presidents will be male, with 100% certainty.
 
2012-11-13 03:05:16 PM

dletter: Aarontology: Its not %100 accurate when the model didn't exist for those thirty years.

They "retroactively" ran it on those numbers.

I don't know how much old poll data you can find going back to the 80s, 70s, etc, (you can probably glean some data from old newspaper articles), but, would be interesting to see how Nate's formulas would have tracked past elections, and how accurate it would be.


When you create a prediction model, it's obviously going to give the right prediction for the data you based it on. You don't get credit for that. You'd have to be a farking retard not to get historical results correct.

The proof of a prediction model is if you can predict the actual future correctly. Which they're currently 0 or 1, and were badly badly wrong.
 
2012-11-13 03:05:48 PM
On a tangentially related note, does anyone remember that old episode of "Alfred Hitchcock Presents" called "The mail-order prophet?" For some reason it jumps to the forefront of my brain whenever anyone talks about a prognostication method that has "predicted presidential elections with 100% success for the last 30 years."
 
2012-11-13 03:06:02 PM

shower_in_my_socks: They missed the first day of Statistics 101 when the first words out of their professor's mouth were "correlation does not prove causation"


How do you know those were the first words out of the professor's mouth?
 
2012-11-13 03:06:03 PM

jjorsett: My back-tested model says all future presidents will be male, with 100% certainty.


mine says no mormon will ever be president and so far so good.
 
2012-11-13 03:06:39 PM
TFA: The model erred, Bickers said, in assuming that a campaign could not shake off such challenging economic factors. One thing lacking, he said, is a good way to model campaign impact.

Yeah. If only there were some way to scientifically survey public opinion. Then you could see what kind of effect campaigns are having, and even better, you can see how much your economic factors are actually influencing voter preferences. It would be even better if you had lots of organizations conducting such scientific surveys - then you could build a statistical model that aggregates them all together to produce a picture of the electorate with a high degree of confidence and a low margin of error.

Sadly, as far as I know, no such thing exists.
 
2012-11-13 03:06:46 PM

sweetmelissa31: "What it means is the president in a sense rose above gravity. He managed to do something that historically doesn't happen"

Obviously Bickers is not a Bayesian.


Meh, Bickers just doesn't know Laplace he belongs is average.
 
2012-11-13 03:06:57 PM

Jackson Herring: Obviously Bickers is not a Bayesian.

I wish I got math jokes :(


When he heard about "posterior probability," he thought they meant calculating probability from one's posterior.

Zing!
 
2012-11-13 03:07:02 PM

lennavan: jjorsett: GAT_00: And running a campaign of "I'm not that guy" doesn't work. The left learned that in 2004. It remains to be seen if the GOP can learn from 2012. You don't get turnout voting against someone, you get turnout by voting FOR someone.

Huh? That's the exact campaign Obama ran against Romney. "That guy's a rich, White, out-of-touch, lawbreaking killer."

That's odd. I searched Obama's Convention speech, you know, the one where he had a chance to speak to tens of millions of people across the nation and the word Romney didn't appear in it at all. Granted, I found it on NPR which we all know is on Obama's payroll, so they probably edited it.

Keep on fightin the good fight man. Hannity would be proud.


The Obama campaign spent hundreds of millions in the early going to demonize Romney, since their guy sure as hell couldn't run on his own record. Keep on drinkin the Kos-aide.
 
2012-11-13 03:07:14 PM
I don't understand why you would try to model people's opinions by assuming their reactions (i.e. what these jokers did vis-a-vis the economy) instead of asking them what their opinion actually is (i.e. a poll). Sure the poll samples may not be representative, but on the aggregate they give you a pretty good guess as Nate Silver proved. Why try to model an outcome using predictor variables when you already have the likely outcome in the form of polls? Oh right, these guys were pandering to the right by making a model say what they wanted it to.
 
2012-11-13 03:07:14 PM

jjorsett: My back-tested model says all future presidents will be male, with 100% certainty.


Does your model include the fact that Howard Taft was a walrus?
 
2012-11-13 03:07:20 PM
Um... every model should have correctly "predicted" (i.e. been consistent with) the last 30 years of results. That's like the first requirement for it being considered a mathematical model at all, otherwise it's just guessing at random.

Whether it's a good model or not depends whether you correlated the right factors with the right weights, there are any number of factors/combinations that correlate well with past data (see "no candidate whose first name has a 'k' has ever lost), it's only the future outcomes that actually count as testing it.
 
2012-11-13 03:07:35 PM

olddeegee: FTA The tool ended up underestimating Obama by 5 to 6 percentage points

Is this his methodology or him?


Black. And. White are. All I see.

in my infancy
red & yellow, pending
 
2012-11-13 03:09:13 PM

jjorsett: The Obama campaign spent hundreds of millions in the early going to demonize Romney, since their guy sure as hell couldn't run on his own record. Keep on drinkin the Kos-aide.


He had long been demonized by his fellow Republicans before Romney got the nomination. How exactly do you blame Obama for that?
 
2012-11-13 03:09:15 PM

Doc Daneeka: TFA: The model erred, Bickers said, in assuming that a campaign could not shake off such challenging economic factors. One thing lacking, he said, is a good way to model campaign impact.

Yeah. If only there were some way to scientifically survey public opinion. Then you could see what kind of effect campaigns are having, and even better, you can see how much your economic factors are actually influencing voter preferences. It would be even better if you had lots of organizations conducting such scientific surveys - then you could build a statistical model that aggregates them all together to produce a picture of the electorate with a high degree of confidence and a low margin of error.

Sadly, as far as I know, no such thing exists.


You know, a thought occured to me. If you ask enough people, random people preferably, you could probably gauge what an electorate is thinking on the whole if you gathered enough individuals until you felt you could realistically extrapolate that group's make up to the entire population.

Nah... it'll never work. Who'd pay for that, anyway?
 
2012-11-13 03:09:51 PM

wildcardjack: Does your model include the fact that Howard Taft was a walrus?


Outlier.
 
2012-11-13 03:10:17 PM
i.imgur.com

/What these professors might look like.
 
2012-11-13 03:10:54 PM

Duke Phillips' Singing Bears: shower_in_my_socks: They missed the first day of Statistics 101 when the first words out of their professor's mouth were "correlation does not prove causation"

How do you know those were the first words out of the professor's mouth?


He borrowed my statistical model which has correctly predicted the first words of every statistics professor over the last 30 years.
 
2012-11-13 03:10:55 PM
Thanks for slightly devaluing my PoliSci degree from CU with your wrongful prediction Mr. Bickers. It's a good thing were not in Italy or I would have you jailed for your failure to predict this accurately.

/This implies my PoliSci degree had worth to be devalued, as I'm sure someone will imply it didn't.
 
2012-11-13 03:11:57 PM
I'm surprise not many have given credit to Bill Clinton for Obama's win. It would be a stretch, but I figured someone would be harping on that.
 
2012-11-13 03:12:04 PM

impaler: Jackson Herring: Obviously Bickers is not a Bayesian.

I wish I got math jokes :(

When he heard about "posterior probability," he thought they meant calculating probability from one's posterior.

Zing!


hahah like a butt!

now THAT is a joke I get
 
2012-11-13 03:12:12 PM

Communist_Manifesto: /This implies my PoliSci degree had worth to be devalued, as I'm sure someone will imply it didn't.


Your predictive model is likely going to be more effective than Bickers'.
 
Displayed 50 of 112 comments

First | « | 1 | 2 | 3 | » | Last | Show all

View Voting Results: Smartest and Funniest


This thread is closed to new comments.

Continue Farking
Submit a Link »






Report