JEM Algo used by Nate Silver's 538 predicts Senate win for Rs.

[jem]

According to 538 as of today... Republicans have a 57.1% chance of winning a majority.

http://fivethirtyeight.com/interactives/senate-forecast/


More interestingly (to me since we already knew that) however is that 538 discloses they use a model to "unskew" the polls.
The effect of their model is to discount the polls which differ from a range of the strong pollsters results. His method gives more weight to pollsters whose polls come in close with the final results of previous races.
So in short his models brings polls back into line with templates created by previous elections.

The JEM algo.... all modeled out. IMO


(note his state model is more complicated and looks reasonable at first glance)
see here...

FiveThirtyEight’s pollster ratings are calculated by analyzing the historical accuracy of each firm’s polls along with its methodology. Accuracy scores are adjusted for the type of election polled, a firm’s sample size, the performance of other polls surveying the same race, and other factors. We also calculate measures of statistical bias in the polls.

http://fivethirtyeight.com/interactives/pollster-ratings/

(I am sure he will say he had the algo first... but I never saw it back when the JEM algo was developed on the skewed polls some which were D plus 11 or more.) Also note this is all in good fun. I realize he has far more experience with stats and polls than I have.




See also... here where he unskews the pollsters in a simple manner.

Perhaps the simplest measure of poll accuracy is how far the poll’s margin was from the actual election result. For instance, if a poll had the Democrat ahead by 10 percentage points and she actually won by 3 points, that would represent a 7-point error. In the table below, I’ve listed polling firms’ average error for elections from 1998 through 2007, and again for the same polling firms for elections from 2008 onward. (About half the polls in our database are from 2008 or later, so this is a logical dividing point.) I’ve restricted the list to the 28 firms with at least 10 polls in both halves of the sample.


http://fivethirtyeight.com/features/how-fivethirtyeight-calculates-pollster-ratings/


===

 

[Tsing Tao]

Oh crap, the Jem algo. Oh well, it was fun while it lasted. Congratulations, Democrats!

 

[Max E.]

Yeah Jem im not so sure id be bringing this algo back after its performance in the last election

 

[Ricter]

Yeah Jem im not so sure id be bringing this algo back after its performance in the last election

Lol, I think we were supposed to have forgotten about that.

 

[jem]

as a professional polling organization we appreciate all comments.
Our investor relations team and our PR depart welcome your direct comments and suggestions. consensus@co2cools.com


the jem algo is a little less confident than 538 at the moment.

right now it has the republicans at a 55.27 % chance of achieving a majority.

But our proprietary JEM Algo Harmonics function is projecting the advantage will move in Republicans favor as we get closer to the election. The harmonics indicator notes that the polls as a group modify their sample sizes as they get closer to the election and their samples almost invariably move in the Rs favor. We have seen the effect starting to kick in as of this week.

We have found 2 primary causes.

1. some polls switch their samples to likely voters
2. other polls modify their samples to conform with previous election templates in the last 2 or 3 weeks before the election. The samples almost always swing to the Rs advantage.

 

[jem]

the polls are starting to swing towards the Rs... surprised?
the JEM Algo Harmoics functions is kicking in.

Have you seen the poll headlines this week.

here is another...

http://www.breitbart.com/Big-Govern...Brown-Jumps-Ahead-In-Polls-For-The-First-Time


SCOTT BROWN JUMPS AHEAD IN POLLS FOR THE FIRST TIME

 

[dbphoenix]

No. It's a midterm election.

 

[Ricter]

No. It's a midterm election.

It's amazing how close it is, in a midterm election (president's second term) which is typically a slam-dunk for the opposite party. Reps have nothing to focus on: criticism of the ACA is falling flat; Benghazi... well there never was "any there, there"; The IRS examinations? Snore. Criticizing Obama's lack of economic stimulus, lol, the reps won't be going there. The jem algo's accuracy this time? "Even a dog's ass gets..."

 

[dbphoenix]

It's amazing how close it is, in a midterm election (president's second term) which is typically a slam-dunk for the opposite party. Reps have nothing to focus on: criticism of the ACA is falling flat; Benghazi... well there never was "any there, there"; The IRS examinations? Snore. Criticizing Obama's lack of economic stimulus, lol, the reps won't be going there. The jem algo's accuracy this time? "Even a dog's ass gets..."

Perhaps not all that amazing. It's much easier to manipulate people if one can keep them fearful, and the rightwing media has elevated that to an art form. And when joblessness is high, wages are low, the recovery is stagnant, ordinary folks are ripe for that sort of thing.

 

[gwb-trading]

Perhaps not all that amazing. It's much easier to manipulate people if one can keep them fearful, and the rightwing media has elevated that to an art form. And when joblessness is high, wages are low, the recovery is stagnant, ordinary folks are ripe for that sort of thing.

Yeah, thanks to the Obama administration joblessness is high, wages are low, the recovery is stagnant, and ordinary folks are ripe for a change.