Df - The post from freeman2 at the top of the very same page as your last post does not read to me as saying Nate Silver is infallible. He talks about how assumptions are made, that social sciences are not that great at predictions, and starts out rubbishing percentage chance predictions.

I know that Ricky has referred to Silver a lot, but I can't see that he's claimed him to be near infallible.

Could you try not putting words into others mouths so you can then caricature them?

Just as you accuse others of treating Silver as a security blanket, it strikes me that the certainty that he is wrong is providing exactly the same comfort to Republicans.

Yeah, apparently DF did not see Ricky take me to task for saying it was not possible to predict with mathematical certainty the outcome of an election.

freeman

Yeah, apparently DF did not see Ricky take me to task for saying it was not possible to predict with mathematical certainty the outcome of an election

Is that what I did? Hree's what you said....

There are too many variables in elections to be making percentage predictions. Silver's model makes a certain set of assumptions and then uses statistical analysis to formulate predictions. However, social science is not like physics--human behavior is not that predictable

I then took you to task for saying that human behaviour is not that predictable....

I agreed that you can't calculate with mathematical precision.... But it is certainly possible to accurately predict the mathematical outcome of an an election (percentage vote or number of seats or electoral votes) . And someone who's made a public prediction about the US election is going to get it mathematically right too. Thereby disproving your claim that it is impossible. There's just too many predictions being made, not for someone to get it bang on. (Thats a mathematical statistics kind of thing) .
Thing is, if you read Silver accurately he isn't making a specifc prediction. He's producing a statistical liklihood. Thats very different.

freeman2 wrote:Yeah, apparently DF did not see Ricky take me to task for saying it was not possible to predict with mathematical certainty the outcome of an election.

Indeed. But you are a lefty, and therefore probably lying. We know what you really mean because... because... that's what all dodgy lefties think, even when they expressly say they don't.

[/sarcasm]

Ricky:

There's just too many predictions being made, not for someone to get it bang on. (Thats a mathematical statistics kind of thing) .

And a stopped clock is right 2X a day.

Percentage predictions are meaningless. They are just odds. As favourites don't always win horse races, and sometimes outsiders do, even carefully constructed odds with large markets around them can be inaccurate in a one-shot race.

And as RJ indicates, a correct result prediction (of national vote or EC outcome) could be plumb luck too.

However, as there are many polls and predictions around, we can also see where they are clustering and what patterns there are. It's not whether one is right, and the others wrong, so much as how out everyone is, and who is most wrong.

In that sense, there's no reason why Nate Silver is more accurate than the RCP, or any other pundit. We'll only know after the fact, and then there'll be a bunch of reasons expounded as to why X was closer than Y.

What I note, however, is not that NS is predicting a win for Obama. It's that most polls and prediction models are doing so. If you completely ignore him, that doesn't really change the aggregate much. He's useful to read because he explains in detail what is going on and comes up with reasons - and questions - that affect what his model predicts. Which means that people are more likely to point to him - not because he's some kind of seer, but because he says stuff of interest.

I don't really know if he's accurate this time (my finger in the air prediction was and still is closer than his, being less like 70 EC votes and more like 20).

I don't care as much about his sports predictions (proper sports should not be that predictable anyway). I can see how he got the UK elections wrong, although he wasn't much further out than our usual uniform swing models (which are known to be flawed, it's just we haven't found a better way). His collation of the polls was pretty good. The polls were wrong, however, because they did not pick the downturn in the Lib Dem vote following the late April surge. In the end, the prediction for the Tory result was very close - and that was the most important part, because it was too low to be a majority but too high for a Labour led coalition to be viable.

Not that it makes much of a difference (because I have forgotten most of it) but I was a math major in college for a few years (and took a couple of upper division
probability classes). If I take a coin and try to figure out what the probability of getting heads ten times in a row, then I can be mathematically certain of that probability. But I just cannot see you be as certain when you're trying to model election probabilties based on polling. You have different selection procedures for obtaining likely voters, you have different response rates, you have different turn-outs with regard to both Republican and Democratic voters from election to election, you don't know how undecideds are going to break, some voters do not like to participate in polls, etc. That is a lot of variables that aren't being controlled for. I think Silver's analysis is interesting and informative, but to say that Obama has a certain mathematical percentage of winning the election is very questionable. Yes, with all else being equal the chances that the polling will be off by more than 2.6 percentage points in Ohio is fairly low. But we don't really understand the underlying variables well enough to start saying that Obama has this percentage chance of winning Ohio. In any case, I like our chances when you average a bunch of legitimate polls that you will somewhat close to the final outcome, but saying that we know exactly what percentage is before the election is a stretch in my opinion

Gallup tracking: Romney 49, Obama 48.

Rasmussen--same.

Gallup's poll comes in at Romney +1. Which is some way closer than the +5 they had before polling was suspended a week ago.

Mittmentum is officially dead.

danivon wrote:I don't care as much about his sports predictions (proper sports should not be that predictable anyway). I can see how he got the UK elections wrong, although he wasn't much further out than our usual uniform swing models (which are known to be flawed, it's just we haven't found a better way). His collation of the polls was pretty good. The polls were wrong, however, because they did not pick the downturn in the Lib Dem vote following the late April surge. In the end, the prediction for the Tory result was very close - and that was the most important part, because it was too low to be a majority but too high for a Labour led coalition to be viable.

NS also, as I posted, weights polls by his discretion. So, if a poll he likes is inaccurate, how much worse does that make his result?

Again, here's the heart of the farce for me: he rates Ohio as 86.8% likely to go for Obama. Nearly all of the polls are within a point or two, some are tied, and all are within the margin of error.

If he were an oddsmaker in Vegas, he'd be going broke.

danivon wrote:Gallup's poll comes in at Romney +1. Which is some way closer than the +5 they had before polling was suspended a week ago.

Mittmentum is officially dead.

Until Wednesday morning, when it's President-elect Mittmentum.

Of course, I love the CNN national poll. It's tied 49/49. So, why do I love it?

Only this poll turns out to be bad news for Barack Obama. Why? According to CNN’s internals of this likely-voter survey, Obama doesn’t come close to tying Romney in one key indicator — independent voters. According to the sample data, Romney leads independents by a whopping twenty-two points, 59/37. How can Romney have a 22-point lead among independents but still only get to a 49-all tie with Obama?

Here’s how:

Among those likely voters, 41% described themselves as Democrats, 29% described themselves as Independents, and 30% described themselves as Republicans.

Yes, that’s a D/R/I of 41/30/29. In 2008, a big turnout election for Democrats, the D/R/I was 39/32/29, and Obama won independents by eight points. Despite Republicans having a five-point edge in this very poll among those “extremely enthusiastic” about voting, the CNN poll has added four points to the Democrats’ advantage in this sample.

Let’s also take a look at the gender gap. In 2008, Obama got a +14 in the gender gap, with a +13 among women and a +1 among men. In this poll, Romney wins men by nine (53/44) and Obama wins women by eight (53/45) for a gender gap advantage of +1 for Romney.

That says one thing and one thing only: Romney wins.

There are polls showing him ahead in IA, OH, WI, and tied in MI and PA. I like his chances.

Doctor Fate wrote:Until Wednesday morning, when it's President-elect Mittmentum.

What was all that earlier about other people being so certain and all?

I think it's likely you will be proved wrong by Tuesday. But I'm not dumb enough to say it will happen.

The other thing, DF, what did the CNN poll control for?

If they polled Registered Voters and controlled for that (and doing so based on registration etc), and then asked a question to sift Likely Voters and recorded the results, it suggests that the differential turnout works well for Obama.

If they controlled for Likely Voters, you have a point - but how did they know they were LVs until they polled them? And what did they do about all the people who were less likely to vote?

I'd be less sanguine if I were you.

danivon wrote:

Doctor Fate wrote:Until Wednesday morning, when it's President-elect Mittmentum.

What was all that earlier about other people being so certain and all?

I think it's likely you will be proved wrong by Tuesday. But I'm not dumb enough to say it will happen.

What happened to leaving invective aside?

I'm not sanguine. I'm sifting through the numbers.

Btw, if you believe that any poll is sifting in such a way to advantage Romney, I'd find that an interesting proof. Knock yourself out. A few pages (eons) ago, I mentioned one poll Obama was leading wherein the pollster managed to remove 4% of the sample from RV to LV. 4%?

Here's part of the reason I'm hopeful: http://hotair.com/archives/2012/11/05/w ... arly-vote/

The difference in early vote turnout, enthusiasm, etc., is palpable. Read all of that column, then tell me it is going to be a walk for Obama. IF he wins, we won't know it until early Wednesday morning. Obama will lose FL, VA, and NC. Then the nail-biting begins. However, with CO trending well, and OH, WI as toss-ups (Obama was there today), and MI, MN and PA as possible upsets, I have every reason to be optimistic.