RE: Why Nate Silver predicts an Obama win. (Full Version)

All Forums >> [Community Discussions] >> Dungeon of Political and Religious Discussion



Message

focalss -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 7:59:25 AM)

I checked the Nate Silver site again and he must have state by state models or applies his models to individual states to get his national prediction.

Without seeing the model I see he has Obama 86.8% chance of winning Ohio, 90% Nevada, 72% VA, 69% CO.  So I request help on explaining how if the Ohio polls are at 2.4%+ Obama that translates to an 86% chance of winning which is why I am critical of Silver's saying 86%.  The little I know of statistics, going with a 2.4% lead does not produce that big of a probability.  You are talking a few standard deviations away. 

I do understand there can be more factors than just polls in a model maybe that explains it.  I have been a little preoccupied the last week with a storm in my neighborhood so wasn't following things that closely and lost track of it.  I am going with the 2.4% quoted by DS.

I saw the news saying that the early voting is not trending overwhelmingly to Obama which early polls would have indicated it would with Obama having a bigger lead earlier in the race.  I also agree likely voters are the key. 

However to DYB, the Nevada situation is the most interesting thing that caught my eye on Silver's site.  Why is Romney behind in Nevada when it has a large Mormon population, bad economy and he ran the Olympics there?  Not a good sign for him and it plays into DYB comment on the other side.  If Obama can get more people out for him where are the Mormons in Nevada for Romney?  Likewise in Ohio, the groundgame is all now in Ohio.  Rove was able to pull out voters in Red districts to go for Bush in 2004 to avoid the Florida 2000 scenario.  Given that big cities trend to the Democrat, I still say nothing changed in the Ohio Red Districts (Red = Republican, Blue = Democratic).  Reportedly Obama has more workers there but are they going to be able to produce and swamp the Ohio Reds?





Moonhead -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 8:03:17 AM)

Aren't only 49% of the Moomins in Nevada allowed to vote for anybody?
(Or is it one of the other flakey American desert pseudo Christian sects who refuse to let women vote?)



Yachtie -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 8:07:28 AM)


quote:

ORIGINAL: focalss

So I request help on explaining how if the Ohio polls are at 2.4%+ Obama that translates to an 86% chance of winning which is why I am critical of Silver's saying 86%. 



Loosely, using the 86% chance, it only means that Obama has an 86% chance of actually hitting that 2.4+% win in Ohio. Statistically, if one utilizes Silver's statistical model, that's not a bet one should go against. But like horse racing, it's been known to happen.



focalss -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 8:30:42 AM)

I guess without the standard deviaition you can't tell but I think in general the polls say that the result has an 85% chance of being what is polled within the margin of error.  That does not to me translate to 85% chance to win, it translates to 86% that the final result will be inside the margin of error.  Romney could win Ohio and the poll statistics can be correct depending on the sd and margin of error.

That is leaving out likely voters, get out the vote efforts etcetera.  Assuming 2.4% + Obama with 0 days left that should translate in normal situations to Obama winning but I don't see how that translates to an 86% chance of win, more to me in the neigborhood of 60-66% if the poll is outside the margin of error.

If the poll had a 2% margin of error lets say and it came in 50 - 49 Obama - Romney then it cannot predict accurately the result unless I misunderstand.



DomYngBlk -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 9:27:43 AM)

quote:

ORIGINAL: DarkSteven


quote:

ORIGINAL: DomYngBlk

DS, I understand all of that and in off year elections I'd say that it is more than true. However, I think that from the early voting we have seen that this is far from a "likely" voter election. So I personally think the data from registered is more likely the real case. We will know more tomorrow night of course


DYB, I'm not following. If the actual voters reflected perfectly the makeup of registered voters, then the registered voters and likely voters samples would be essentially the same. The very fact that they are different means that the registered voters will not vote with equal Dem/Rep representation.

Unless you think that the methodology to determine what are "likely" voters is flawed.

What is your rationale for saying that this election will be different? I get the feeling that it's due to the makeup of the early voters - true?


Yes the makeup will be quite different. I think that the actual voters this time will comprise far more accurately a look at the registered voters than would normally be the case in an off year election.

There is a reason that the polls in all the swing states were heavily for Obama. The ground game they have reflects that strategy. Get the "registered" or hard to get to the polls voter out to the polls first then work on Election day to make sure your "likely" voter is in the mix.

If you also add in the swing states having been told they are "more" important this year there will likely be high turnouts in all of them. All of that breaks well for Democrats and Obama.

I am not saying Silver is wrong. Hardly. Stats don't lie.



tazzygirl -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 9:55:00 AM)


quote:

ORIGINAL: TheHeretic

The straight fundamentals (the shit economy, which is rated as the most important concern, and the tendency of undecided voters to break to the challenger) say Obama will lose, except for the elephant in the room, and nobody will know until Tuesday night how that plays itself out.




It was addressed.

Others argue that undecided voters tend to break against the incumbent, in this case Mr. Obama. But this has also not really been true in recent elections. In some states, also, Mr. Obama is at 50 percent of the vote in the polling average, or close to it, meaning that he wouldn’t need very many undecided voters to win.



DomYngBlk -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 10:08:10 AM)

quote:

ORIGINAL: focalss

I checked the Nate Silver site again and he must have state by state models or applies his models to individual states to get his national prediction.

Without seeing the model I see he has Obama 86.8% chance of winning Ohio, 90% Nevada, 72% VA, 69% CO.  So I request help on explaining how if the Ohio polls are at 2.4%+ Obama that translates to an 86% chance of winning which is why I am critical of Silver's saying 86%.  The little I know of statistics, going with a 2.4% lead does not produce that big of a probability.  You are talking a few standard deviations away. 

I do understand there can be more factors than just polls in a model maybe that explains it.  I have been a little preoccupied the last week with a storm in my neighborhood so wasn't following things that closely and lost track of it.  I am going with the 2.4% quoted by DS.

I saw the news saying that the early voting is not trending overwhelmingly to Obama which early polls would have indicated it would with Obama having a bigger lead earlier in the race.  I also agree likely voters are the key. 

However to DYB, the Nevada situation is the most interesting thing that caught my eye on Silver's site.  Why is Romney behind in Nevada when it has a large Mormon population, bad economy and he ran the Olympics there?  Not a good sign for him and it plays into DYB comment on the other side.  If Obama can get more people out for him where are the Mormons in Nevada for Romney?  Likewise in Ohio, the groundgame is all now in Ohio.  Rove was able to pull out voters in Red districts to go for Bush in 2004 to avoid the Florida 2000 scenario.  Given that big cities trend to the Democrat, I still say nothing changed in the Ohio Red Districts (Red = Republican, Blue = Democratic).  Reportedly Obama has more workers there but are they going to be able to produce and swamp the Ohio Reds?





2004 the turnout in northeast ohio among democrats was paltry. That is why Kerry lost Ohio. Obama looks to win ohio with NE Ohio, toledo, dayton, and southeast ohio plus the cities of columbus and cincy. Key it getting people to the polls....hence that many offices

That effort has been duplicated in each of the swing states. The republicans have no such effort. Rove and Romney are spraying ad money around but don't have the people to get into the suburbs and rural areas they have to dominate.



Yachtie -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 10:11:59 AM)

This is tangential to Silver's statistics -

The demographic threat to the Republican Party grows out of the fact that every four years the electorate becomes roughly two percent less white and two percent more minority, primarily as a result of the increase in the Hispanic and Asian-American populations and the relatively low birth rate among whites. By my computation, this translates into a modest 0.85 percentage point gain for Democrats and 0.85 percentage point loss for Republicans every four years. In other words, the changing composition of the electorate gives Democrats an additional built-in advantage of 1.7 percentage points every four years.

The democrats may not get it this round, or even perhaps the next (2016), but statistically the US is doomed to the (left of center minimum) ideology of the left as more immigrants align with the left than the right.






tazzygirl -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 10:13:30 AM)

Ya might want to think on that for a while. Swirl it around. Chew on it, as my dad would have said. And consider why that is.



Yachtie -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 10:19:29 AM)


quote:

ORIGINAL: tazzygirl

Ya might want to think on that for a while. Swirl it around. Chew on it, as my dad would have said. And consider why that is.


'Why' was of no concern to my comment. Only that the statistic says it is heading that direction.



tazzygirl -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 10:26:43 AM)

lol... of course it is. For myself, Im laughing over the upset so many seem to have over the potential of whites being a "minority"... enough of one to eventually be discounted.



truckinslave -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 11:22:22 AM)

Nate Silver is no doubt a helluva statistician, a helluva numbers guy.
But it's still GIGO, still based on assumptions of who will show up; the results are still controlled by the modeler.
(Sorta reminds you of that non-existent anthropomorphic global warming in that regard, uh: modelers, indeed).
Michael Barone, on the other hand, is perhaps the most knowledgeable man in the world concerning American elections and the American electorate. He went out on a limb recently.

If Pa goes Romney, it's over.
If Va goes Obamao, it's probably over (unless Romney carries Pa)
As Ohio goes, so it goes (unless Romney carries Pa)

Those other states are just wasting money opening the polls.



Hillwilliam -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 11:25:06 AM)

Truckin. Long time no see. Welcome back just in time for the festivities and the whining afterwards for whichever side loses.



cloudboy -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 11:27:59 AM)

quote:

Without seeing the model I see he has Obama 86.8% chance of winning Ohio, 90% Nevada, 72% VA, 69% CO. So I request help on explaining how if the Ohio polls are at 2.4%+ Obama that translates to an 86% chance of winning which is why I am critical of Silver's saying 86%. The little I know of statistics, going with a 2.4% lead does not produce that big of a probability.


Without going back and reading all his material, a percentage point lead has a particular election result based the past results. So in the past a 1 pt lead might have been 56% success rate of winning. Each pt up translates into a much higher past election success rate. Anything nearing a 3 pt lead equates to about 90% chance of winning. So, as the percentage pt lead increases by 1-2-3 pts, the outcome election prediction jumps in a much higher proportion.

For instance an 8 PT has a 100% success rate.

Not sure if that answer your question.

quote:

If Pa goes Romney, it's over.
If Va goes Obamao, it's probably over (unless Romney carries Pa)
As Ohio goes, so it goes (unless Romney carries Pa)


Keep dreaming Mr. Truck.



tazzygirl -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 11:30:48 AM)

PA wont go Romney.



focalss -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 2:58:10 PM)

I looked at the Barone predictions and hope he has some good net under the branches when they go.  He is giving every single toss up to Romney.

Back to the 2.4% question.  There is a margin of error and standard deviation in there somewhere.  Maybe thats the way it is, but to me a 2.4% lead doesn't translate to 86%.  I have been looking for some statististics information and can't get what I want nor can I find out the details of the polls, error and std dev.

The polls I just looked at have a 1% difference.  The margins of error are around 2-3%  That means they can't predict that much except it will be a close race.  So with a 1% poll from 270 to Win that I follow, and some polls I haven't seen that say there is a bigger spread or the firewall worked I can't really respond except that they all seem to me to be within the margin of error and again how Silver gets 86% probability of an Obama win is what I am not understanding.

Where are the Mormons in Nevada?  Have they abandoned Romney?



mnottertail -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 3:03:14 PM)

If the 86% is a probability it is statistically speaking (based on some uncertainties and some (for the purpose here we will say known) known variables.

And there is in his calcs, a 14% uncertainty of Obama's election. 

Again if that is a probability.



slvemike4u -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 3:14:44 PM)

This thread is no fun at all.
Nothing but percentages and friggen numbers
I'm getting a fucking headache...which is why I said in my first post that the guy is a genius,he works with numbers...and doesn't get a headache [:)]



tazzygirl -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 4:00:42 PM)

You are looking at the polling numbers.... not the probability of him winning that state election.

They are two completely different numbers.



DomKen -> RE: Why Nate Silver predicts an Obama win. (11/5/2012 4:43:53 PM)

The idea is pretty simple.

One poll may be an aberation. A bunch of polls all with the same result, or nearly the same result, increases the confidence that those polls are correct. Mr. Silver says Obama has an 85+% chance of winning because looking at the recent polling in the battleground states he has that chance, based on the historical accuracy of those polls, of winning enough states to reach 270 electoral votes. The remaining 15% is basically the chance that all the pollsters have misidentified this years electorate or have some other endemic sampling error.

You can look at this a lot like baseball. If a hitter hits .325 for a single season you'd be pretty foolish to count on him hitting nearly that the next season. However if he's hit .325 every season for the past 5 years then there is a reasonable basis to assume he will hit in that area again (assuming no injuries or other factors).



Page: <<   < prev  1 [2] 3 4   next >   >>

Valid CSS!




Collarchat.com © 2025
Terms of Service Privacy Policy Spam Policy
4.699707E-02