Nate Silver is getting hammered by the right for daring to use math-based math rather than faith-based math. In his defense, I thought it might not hurt to take a quick review, for non-math types, of the mathy side of things.

This will be old hat for some of you. No quiz, I promise. More after the orange ampersands' act of public indecency.

Update: Uh-oh, Community Spotlight! Now I have to fix my typos!

Polling

Suppose you've got a cubic yard of M&Ms of various colors, and you want to know how many are blue. The only way to know for sure is to count them all, kinda like that study of the four thousand potholes in Blackburn, Lancashire that John Lennon mentions.

But you can make a statistical assumption - that if you look at a smaller sample, and count the number of blue M&Ms in that, you can derive an estimate of the total, as long as you're willing to allow for some wiggle room. In a sense, the statistics of polling is all about defining what "wiggle room" means.

That wiggle room comes in two flavors: margin of error, and level of confidence. Here's exactly what that means.

AP recently issued the results of a poll showing that 18% of Americans think Obama is a Muslim. Up at the top of the first page, in tiny print, we're told: the sample size is 1037, the margin of error is ±3.04, and the confidence is 95%.

What that means, in English, is this: "We can't ask everybody, but we asked enough people (1037) that, according to the standard mathematics of sampling, we can conclude that there is a 95% chance that the actual number we'd get if we did ask everybody would be somewhere between 18-3.04 and 18+3.04."

These three numbers -- sample size, margin of error, and level of confidence -- are all related, but in a way that requires calculus rather than algebra to define, so I won't drag you through it. The basic trends, though, are what you'd expect: increase the sample size, and you lower the margin of error; to increase the level of confidence, you need to increase the sample size; given the margin of error you want, you can calculate the sample size you need; geekery geekery geekery abounds.

These three numbers are the Patty, Maxine, and Laverne of polling; each only makes sense in the context provided by the other two. But there is also a convention that says, if a confidence level is not mentioned, it's the default value of 95%.

What 95% also means is that even the bestest of the bestest poll is going to get it wrong one time out of twenty. Even the best poll throws a piston every now and then. It's the price we pay for not having to count all the M&Ms. And it's also the reason statisticians tell you never to take any single poll as The Truth, because you might be looking at one of those thrown pistons.

Sampling Error

We've got a built-in assumption here that the sample we're counting really reflects the whole. And this is where polling methodology matters. It's easy to, say, shake up a cubic yard of M&Ms (if you have ready access to a fork lift), and then grab a sample. With people, it's harder. How do you make sure your sample really does match your general population?

This is where variables -- human poster versus robot, land lines versus cell phones, time of day, etc. etc. -- come into play. Generally, the sample population isn't going to match the general population exactly, and polling firms have to apply corrections, based on their own assumptions of what the general population is actually like. Different polling firms make different assumptions.

Those assumptions, along with the variables I mentioned above, give each polling firm a "house effect" -- that is, a bias. If you know that, for example, Rasmussen is always going to lean two percent more Republican than other polls, you can do one of two things. You can decide to chuck Rasmussen results altogether, or -- Nate's approach -- you can estimate the house bias and subtract it for any given poll result. The same holds true for left-leaning polls.

So what Nate's doing when he figures out, in any given day, how things are in, say, Ohio, he's not averaging the direct poll results, as the Republican talking points suggest he is, but he's calculating the results based on the corrected poll results, after the house effect has been estimated and removed from each poll.

Why Nate Silver is The Bomb

So then there's a pile of states, each of them having a calculated percent of Obama support and a margin of error. Once you've got that, and throw in some basic stuff about the Gaussian bell curve, you can calculate the odds that Obama wins a given state.

How does that stack of probabilities turn into a single probability figure for the whole country? The answer there is to use a technique that comes from -- not making this up -- the bomb makers at Los Alamos. Faced by a calculation that was too tricky to work out via pure math on the chalkboard, they tried another approach: doing repeated simulations, with a certain amount of randomness thrown in, and then averaging the results. The more simulations you try, the smaller the margin of error in your final result. Given that each calculation includes, by design, a slice of random chance, In 1946 the Los Alamos mathematician Stanislaw Ulam named it the Monte Carlo method, after the casinos there.

So Silver runs an experiment. He randomly generates election results for each state consistent with the polling data, counts the resulting EVs for each state, and checks to see who won. Then he does it again, with different results that are both random and consistent with the polling data. And again and again, until he's got ten thousand trial elections, all different. The percentage of those trials that Obama wins? That's Silver's top-level number. It won't be right on the nose, but its margin of error can be calculated. (There is a related level of confidence here too - the more trials, the less blurry the result.)

He can also check those ten thousand trials for questions like: did Romney win the popular vote but lose the electoral college? Would this result have been different if Ohio went the other way (i.e. was Ohio the tipping point)? Will Obama lose any states he won in 2008? And by checking how many of those ten thousand trials fit the criterion, he can estimate what their probabilities are.

There's another level

The way I've described it presumes that every state is independent -- that is, that the way Wisconsin goes has nothing to do with the way Ohio goes, or vice versa. But there is a correlation. In the real world, the odds of, say, Florida going blue are going to be related to the odds that Ohio goes blue; a red Ohio means a blue Florida is less likely than if there were a blue Ohio.

So Nate also includes for each state factors like this, although I don't think he's described it in detail. What matters is that it isn't being done subjectively but numerically; it's not something by definition works for or against Democrats.

Nate also has some guides on how the undecided voters are going to finally pull the proverbial lever, based on a variety of political and economic indicators.

The Republican Attack

Since the results are showing a pretty solid probability of an Obama win -- by definition, the methodology can at best generate probabilities -- various faith-based math types are trying to attack Silver's methodology as being inherently biased somehow. This seems to be the year that the GOP taught its low-information voters the word "oversampling," for example. They're also claiming some ridiculous numbers for how independent voters are breaking. And they are in turn attacking Nate for the assumptions he's making on how the undecided are going to land.

But, again, the assumptions Nate makes aren't partisan, but based on analysis of previous elections, and he's gone to some trouble over the last few months to spell these assumptions out.

But think like a Republican for a moment. You've spent four years complaining about the Marxist Kenyan Welfare-State Food-Stamp Socialist in the White House. Along comes Mister Math Guy saying: it's gonna be four more years! Now, which is easier to do at that moment - accept hard reality about the President You Hate So Much, or just attack the math guy, 'cuz you were never really all that fond of math guys in the first place?

Tags

EMAIL TO A FRIEND X
You must add at least one tag to this diary before publishing it.

Add keywords that describe this diary. Separate multiple keywords with commas.
Tagging tips - Search For Tags - Browse For Tags

?

More Tagging tips:

A tag is a way to search for this diary. If someone is searching for "Barack Obama," is this a diary they'd be trying to find?

Use a person's full name, without any title. Senator Obama may become President Obama, and Michelle Obama might run for office.

If your diary covers an election or elected official, use election tags, which are generally the state abbreviation followed by the office. CA-01 is the first district House seat. CA-Sen covers both senate races. NY-GOV covers the New York governor's race.

Tags do not compound: that is, "education reform" is a completely different tag from "education". A tag like "reform" alone is probably not meaningful.

Consider if one or more of these tags fits your diary: Civil Rights, Community, Congress, Culture, Economy, Education, Elections, Energy, Environment, Health Care, International, Labor, Law, Media, Meta, National Security, Science, Transportation, or White House. If your diary is specific to a state, consider adding the state (California, Texas, etc). Keep in mind, though, that there are many wonderful and important diaries that don't fit in any of these tags. Don't worry if yours doesn't.

You can add a private note to this diary when hotlisting it:
Are you sure you want to remove this diary from your hotlist?
Are you sure you want to remove your recommendation? You can only recommend a diary once, so you will not be able to re-recommend it afterwards.
Rescue this diary, and add a note:
Are you sure you want to remove this diary from Rescue?
Choose where to republish this diary. The diary will be added to the queue for that group. Publish it from the queue to make it appear.

You must be a member of a group to use this feature.

Add a quick update to your diary without changing the diary itself:
Are you sure you want to remove this diary?
 Unpublish Diary (The diary will be removed from the site and returned to your drafts for further editing.) Delete Diary (The diary will be removed.)
Are you sure you want to save these changes to the published diary?

Comment Preferences

• Nate's genius is calculating the house effect...(48+ / 0-)

of the various pollsters and accounting for it in his calculations.

In statistics, we refer to "garbage in, garbage out" and that is essentially what we get from the poll aggregator sites. All polls are treated with the same level of credibility and it just gets tossed into one big pile to form the average. There is some value in this as it shows areas of consensus and trendlines, but the averages are garbage.

Nate smooths out bias - based on historical house effects - and makes even GOP-leaning groups informative.

• That's an important point.(16+ / 0-)

It means that, rather than subjectively deciding what is or isn't a poll to include in his amalgamation, he's instead including them all but with the relevant, and neutral, correction.

It's one less way to put his methodological thumb on the scale.

Pro-Obama.

[ Parent ]

• I like Nate, but I prefer Sam Wang...(22+ / 0-)

There are three differences between them.

1) While Nate "corrects" national and state polls using calculated house effects, Sam takes only state polls, and uses the median value, whic automatically excludes all of the most partisan polls from both sides.

2) Nate also throws in economic factors that aren't really measurable.

3) Sam uses a polynomial function to calculate the probability of each of the 2.3 quadrillion (2 to the 51st power) combinations of state results, and summs the probabilities for all possible EV totals.

The result is a much more exact solution.  The fact that Sam's calculation gives a higher probability of winning (96 to 99% now) is pure gravy.

The Scout Law (trustworthy, loyal, helpful...) is a GREAT liberal manifesto.

[ Parent ]

• does nate figure in ground game/enthusiasm?(1+ / 0-)
Recommended by:
Lujane

if so, how?

Get your facts first, then you can distort them as you please. -- Mark Twain

[ Parent ]

• I can't quite get #3 to my satisfaction, and(2+ / 0-)
Recommended by:

I would love to see more about Wang showing 96%-99% now. I'm having trouble finding that info on the Princeton Consortium web page. The sixth link down, titled "Probability of Obama Re-election: Random Drift 96%, Bayesian Prediction 99%" is dead. Sigh. What is the difference between the two numbers?

Also, back to #3 above, can you detail that a bit more? Or show me where Wang does? I'd love to find an explanation similar to the one the diarist uses.

The man who moves a mountain begins by moving away small stones. -Confucius

[ Parent ]

• Wang got zinged by Sandy(3+ / 0-)
Recommended by:

He posted a diary here yesterday -- either he or his server or both are in the affected area of NYC. So he's somewhat crippled at the moment, and that may be why you're having a problem with his website. Keep trying, or find his diary ("mindgeek" I think) and ask him directly.

• I did see that diary. I'm just reluctant(2+ / 0-)
Recommended by:

(for whatever reason) to bother him directly with such a seemingly trivial thing. I was hoping someone else would have figured this out already. There is so much information it is kind of confusing.

The man who moves a mountain begins by moving away small stones. -Confucius

[ Parent ]

• he likes commenters(1+ / 0-)
Recommended by:
smartdemmg

and he's taking comments at his face book page right now.

If transferring wealth to the top 2% creates jobs, wouldn't we be swimming in jobs right now? (-9.75 / -9.05)

[ Parent ]

• Random Drift vs. Bayesian Prediction(9+ / 0-)

Statistically, the race won't stand still.  There are two different models for how this could change.

The random drift model says that Obama's chances of winning might increase or decrease randomly over time, even with no significant event (like a debate)  The spread increases over time, from the current starting point.  The prediction is that in 96% of the cases, by random drift, Obama would win.

The Bayesian Prediction says that over the past six months or so, there's an average margin of victory for one candidate or the other.  Individual events might disturb this equilibrium one way or the other, but as time passes, the race settles back towards that average.  The situation was disturbed by several events prior to the debates, the DNC bounce and the "47%" comments, and Obama was above his long-term average.  Then he "lost" debate 1, and Romney got a boost.  Since the VP debate arrested Obama's fall, he's been drifting back upwards towards his long-term average of electoral votes and popular vote margin.  Since he hasn't reached it yet, the Bayesian prediction assumes that this trend will continue.  The current prediction is 99%.

As we get closer to the election, the two numbers will converge to the point that Sam can provide a more definite prediction.

The Scout Law (trustworthy, loyal, helpful...) is a GREAT liberal manifesto.

[ Parent ]

• So basically,(0+ / 0-)

Dr. Wang's model argues that in order for Romney to win, something extremely bad for Obama needs to happen within the next 5 days - something that'd shift all the state polls by 2.5 points (and quickly).  Like what happened after the first debate.  Perhaps something even more significant.

I really don't see that happening.
Mitt's time has run out.

• Thank you for this superb elaboration. I would(0+ / 0-)

still like to ask another question as a followup, if I may. I might not ask this with as much aplomb as you responded, but please bear with my groping for the words/ideas. And please don't feel compelled to change or 'dumb down' your style of response. I think you nailed the technical sophistication of the language perfectly. I just have to absorb it.

When I think about what you described as random drift, I consider in my mind that there is some immediate number (let's say 96%) that represents the "now cast" or something akin to it, namely Obama's chances of winning as things stand or appear today. So then, looking forward to what might change in the future (albeit just a few days), when I apply random drift I expect the spread to be symmetric about the immediate number representing today. So I would see 96% +/- some percentage spread, let's say 2%.

Is it true that the spread with random drift will always be symmetric? If so, then I would imagine that there are no calculations necessary to ascertain that "in 96% of the cases, by random drift, Obama would win." All one needs to know is the current number, namely 96%.

Or maybe it's more complicated than that because of the electoral college. I guess I'm not perfectly clear what the random drift specifically applies to. Is it more like random drift of the outcome percentage? Or more like random drift of something more fundamental (like state-by-state votes expected), which then requires a derivative calculation to determine outcome scenarios.

The man who moves a mountain begins by moving away small stones. -Confucius

[ Parent ]

• I shouldn't link to it but if (22+ / 0-)

you are interested in the right's "rational" repudiation of Silver, there's actually a lucid article there by this guy Dan McLaughlin. It basically says that Silver's model is based on prior experience, but that relying too much on prior experience is a mistake. His thesis, and the one I think most of our nemeses on the right believe, is that Independents are being undercounted. They maintain that Obama is losing amongst Independents by double digits, and that no one has ever been elected who loses Independents that bad. I don't know if that is correct, but I did look at some crosstabs on different polls out of OH this week and from what I can tell, Obama and Romney are basically tied amongst Independents. I think the right's main argument right now is either based on wishful thinking or on national polls; in any event they are convinced - convinced - that Romney is going to win. Every right wing blog is saying that if Obama wins the only explanation could be fraud. Go check it out if you're interested, it's fascinating; the Dark Side beckons!

• I'm confused, doc(31+ / 0-)

You say they contend

relying too much on prior experience is a mistake
and polling results are suspect because
no one has ever been elected who loses Independents that bad
Doesn't point (a) disqualify point (b)?

Am I right, or am I right? - The Singing Detective

[ Parent ]

• I just call em like I see em.(14+ / 0-)

Go over there and see if you can do a better job of deciphering their secret form of logic. This is the meme du jour of the Tea Party right.

• You're not confused. Let me explain:(32+ / 0-)

1. Math says (R)money is losing
2. Obama is the Anti-Christ
3. (R)money is winning!

We must use what we have
to build what we need. -Adrienne Rich

[ Parent ]

• Basic difference(7+ / 0-)

between statistical models like Silver's and mechanistic models like weather forecast models. Silver assumes historical correlation patterns (adjusted, if appropriate, by things like demographic change) but correlations themselves are really derived statistically and can and do change. Weather forecasts (and similar) OTOH use known physical and chemical mechanisms. F = ma for whatever m and whatever a. The enthalpy of vaporization for water at STP is 2.5 MJ/kg, now, last year, and a million years ago; here, over Australia, over Pangaea way back in the day.

Not to criticize Silver, in fact IMO he's got one of the better models for this purpose. But it's at heart a different kind of model, and conflating it with mechanistic models helps not at all.

I also wonder about the fixation with polls, unless you're in the campaign itself. The only one that matters happens next Tuesday. Unless, of course, the polls are driving behavior. If so, that may (or may not) be a social good, but it is definitely a polling bad.

• Not any more.(1+ / 0-)
Recommended by:
Wreck Smurfy

Modern stat systems are flooded with numbers.

Weighing the values of the sources is as important as the mechanical calculations -- heat pump in hydrology, whatever.

Econometric models, particularly, show up with conflicting "cause" sources.

--- WORD OF GOD MODEL ---

Yep....

Take Rasmussen and correct 4%-for-bias.

If it's a small (R) state then bump to 6%-8%-10% depending on how bad the (R) candidate needs the goosing.

We changed our mailer selections for Florida when WOG told us Obama had flipped ahead. The game, then, got to be poaching and nailing down those fewer-but-surer Election Day tallies.

This has been web-site-ized -- the noras.html page:

• I have no idea what you just said(5+ / 0-)

in that first paragraph, but knowing that there are smarter people than me here posting on DK, makes me feel much better about the Election :)  I'm addicted to Nate's page, check it twice a day but I don't understand how he gets the numbers he gets.

It's not often I have to grab a dictionary and /or google terms (but I'm not a mathematics oriented carbon unit), but I had to for this series of comments.  Thank you for making me learn something new today!  I'm one of those that totally freaked out in algebra because there were LETTERS in the equations - what the heck are letters doing there - this is not English class!

My son, on the other hand, was instrumental in helping me pass my college algebra when I had to take it 10 years ago (30 years after my middle school class) because he can figure out equations to word problems via my reading them to him long distance - they just make me burst into tears - I don't CARE about time and distance ratios - just let me get on the train and go somewhere!  He had AP Calc in high school & passed with good enough grades to get full college credit.

He has to use his math everyday calculating medication dosages based on weight and processing averages based on critters' metabolic abnormalities and a whole bunch of other variables (hey!  I remembered they're called variables!)  Guess it's a good thing I didn't pass along my mathematical disabilities - he'd be a really crappy vet otherwise.

Thank you all for trying to explain how the polls work, makes me feel better after hearing idiots on the radio saying all the polls have to be skewed.

"Focusing your life solely on making a buck shows a certain poverty of ambition. It asks too little of yourself. Because it's only when you hitch your wagon to something larger than yourself that you realize your true potential." - Barack Obama

[ Parent ]

• Sorry about that(3+ / 0-)
Recommended by:
celdd, Ricochet67, Native Light

Hopefully I won't make things worse, as this post isn't about modeling details.

Statisitical models (such as economic or political models) rely on correlations to describe behavior and make predictions. That's only possible if you have a large enough sample size. Fancy statistical models (PCA, eg) may attempt to determine the contributions of underlying mechanisms, but those aren't necessary for prediction and in any case have to assume mechanisms. It's a truism that you cannot infer causation from correlation, and Silver (to use him as an example) doesn't claim that. He says: IF behavior this election cycle is similar to that of previous cycles then the likelihood candidate X wins the election is Y%. That's a big "if", but probably a sound one (though not ironclad). Any number of things could nullify that "if", and members of one party seem to be putting their hopes in election fraud.

Statistical models also require a raft of assumptions, without which it becomes hopeless. A couple of crucial ones are stationarity and the normal distribution. These are frequently good assumptions, but not always. Those übersophisticated economic models with literally trillions of dollars riding on them all went kersplat in 2008 when stationarity failed and distributions suddenly became anything but normal ("in a crisis all correlations go to 1"). We, of course, got to pick up the pieces.

A mechanistic model is different. F(orce) = m(ass) times a(cceleration) whether you're talking about weather systems, racecars, atomic motion, galaxies sailing through space. There are no correlations, no qualifying IF here: F does not have a 70.4% of equalling m times a when it rained yesterday but 92.3% if it rains tomorrow. The enthalpy of vaporization of water (the energy change associated with condensation or evaporation) is a property of water and doesn't depend on previous elections or whether we feel that the DOW will hit 36000 or that America is in decline. Physical and chemical behavior are universal, and mechanistic models use these behavioral "laws" to build the model from the ground up.

• it's a very sound "if" actually.(2+ / 0-)
Recommended by:
celdd, Native Light
IF behavior this election cycle is similar to that of previous cycles then the likelihood candidate X wins the election is Y%.
individual behavior and thought processes do not change all that radically, in the short run, absent some compelling, extraordinary event (say, a "near-death" experience, for example). as a result of this truth, using the results of the 2008 election, as a predictor for the current election, is a perfectly legitimate method, for statistical analysis. in fact, there are entire predictive models, based on historical data, that have been in use for decades. multiple linear regression and correlation analysis is the first one that comes to mind, which is pretty close (if i understand his methodology correctly) to what mr. silver is doing.

for anyone to seriously argue that, by using the 2008 results, mr. silver's method is rendered invalid, means that they either have no clue how predictive statistical analysis actually works (very likely), or they're just full of shit (also very likely). in fact, i'd say the likelyhood of both is damn close to 100%.

• OT but the brain changes(0+ / 0-)

so what you found difficult in childhood might not be so bad now. I really started learning math at 30,   and it became enjoyable instead of the misery it was as a kid.

The founding fathers knew of the mutually corrupting influences of Church and state, wisely sending them to opposite corners.

[ Parent ]

• Maybe that's why I did better(0+ / 0-)

on my re-taking of college alebra - actually aced the class and passed the exit exam with 100%.

After about a dozen midnight crying jags on the phone with my son because I just didn't GET it.

He gave me hints and tips and just said practice, practice, practice - our homework was optional (unlike high school).  I was the only one who did every single problem in the book.  I went to the lab every Saturday and for hours would do problems.  Since we only met once a week, we were doing 3 chapters a week.  The minimum was 200 problems, one weekend I did over 500.

But he was right - I started getting it.  I started questioning in class when I didn't understand how our prof arrived at an answer (he was only a little older than my son - very strange to have somebody half my age teaching math!)  He was patient and often stayed after class to help the non-traditional (i.e., old) students like me.  Plus he was a UF Gator and so was my son - I think he felt it necessary to help a fellow Gator's thick headed mom get through it :)

I always did well in Geometry, but Algebra and Algebra II / Trig were a disaster in high school - never got to Calc at all and looking at my son's book was boggling.  I had no idea how he could read the questions and figure out what had to be done to solve them.

"Focusing your life solely on making a buck shows a certain poverty of ambition. It asks too little of yourself. Because it's only when you hitch your wagon to something larger than yourself that you realize your true potential." - Barack Obama

[ Parent ]

• The Mormon "One Mighty and Strong" messiah(2+ / 0-)

fulfills their White Horse Prophecy.

Problem is, this is a grossly incompetent misreading of The Revelation of St. John the Divine, Chapter 6.

This figure is the conqueror.

2. And I saw, and behold a white horse: and he that sat on him had a bow; and a crown was given unto him: and he went forth conquering, and to conquer.
The Lamb has opened a Seal, but it is Satan who sends forth the conqueror. Their actions alternate.

And in the grand poetry of The Revelation, he is the final Anti-Christ. The one who has 42 months to lay waste. The rider on a white horse of Chapter 6, the first of the Four Horsemen of the Apocalypse.

I do feel sympathy for dedicated LDS Saints, who want to believe every work spoken or written by Joseph Smith. He did not earn their devotion.

• the apocalypse of st john is some mad(1+ / 0-)
Recommended by:
bontemps2012

freaky crazy stuff!

If transferring wealth to the top 2% creates jobs, wouldn't we be swimming in jobs right now? (-9.75 / -9.05)

[ Parent ]

• Actually...(0+ / 0-)

They compliment the same point.

Winning independents in 2008, 2004, or 1920 doesn't correspond to winning independents in 2012, because they're not the same number of independents. If Romney is up 66% to 33%, it makes a very difference if that group is 3 million voters big, or 3 voters big.

Besides, these days, everybody's mostly decided. How many "independents" can there really be?

“Never attribute to malice that which can be adequately explained by stupidity.” - Hanlon's Razor

[ Parent ]

• Quite a few(3+ / 0-)
Recommended by:
InfraredRoses, Leap Year, 123man

Because lots of tea baggers call themselves "independent" but actually are solid R voters.

I suspect that's the reason that A: Democrats are outnumbering Republicans by a significant margin and B: Romney is winning "independents" and C: math challenged Republicans are in for a big surprise because they are essentially counting tea baggers twice when they correct for the Democratic skew relative to Republicans

• i don't think "independent" is the same thing(0+ / 0-)

thing as "undecided". one can be "independent" and have already decided. in fact, most people who claim to be "independent", vote the same party every election, so there are very few actual "independent" voters. but it makes them feel superior to call themselves that.

anyone who is truly "undecided" at this point, has either been living in a cave for the past year, or is a moron, and probably, if there was a god, wouldn't be allowed to vote.

• or seriously bi-polar(0+ / 0-)
anyone who is truly "undecided" at this point, has either been living in a cave for the past year, or is a moron

If transferring wealth to the top 2% creates jobs, wouldn't we be swimming in jobs right now? (-9.75 / -9.05)

[ Parent ]

• That makes no sense.(15+ / 0-)

Nate isn't undercounting independents...hes not counting anyone. He's not in the business of weighting or determining party ID for likely voter models....that's for the pollsters to determine. He's simply modeling multiple points of data from polls others have conducted. The only place where Nate uses prior experience is to correct house effects of pollsters. If the pollster has a demonstrated history (and in 2010 as well as 2008 and before) of putting their thumb on the scale, he will correct for that - and appropriately so.

So this is hardly a repudiation, rather its yet another lame attempt to "unskew" the polls. What McLaughlin is saying in essence is that the polls are all wrong because polls show independents going hard for Romney and the polls undercount them. In other words, the polls are all jacked up with the exception of that subsection of the poll where Romney has an advantage. That part is dead on. Uh huh.

• What I find interesting is(22+ / 0-)

their level of emotion. It's equal to ours, and completely the opposite on every single issue. They are angry that Obama is president, and that he might be reelected. They are angry that people are attributing Sandy to climate change. They think that Obama should be impeached over Benghazi. They think that Axelrod and Plouffe determine the employment numbers based on Obama's political needs, and send them over to the Dept. of Labor a week in advance.

They are not going to accept an Obama victory.

• Their lives are based on anger.(12+ / 0-)

The world is constantly conspiring against them to take what they are entitled to. They are being left behind and need to have a system in place that rewards them with advantages based not on their ability, but on their membership in the lucky sperm club.

• My life was based on anger when(19+ / 0-)

Bush was president. It will again be based on anger if Romney wins.

• But our anger is way more rational than theirs(5+ / 0-)

snark or maybe not.

My AIDS/LifeCycle homepage: http://www.tofighthiv.org/...

[ Parent ]

• But what if after the election(11+ / 0-)

Mitt implements a single payer health care system, favors a clean energy and infrastructure building approach to our future and favors a constitutional amendment to create marriage for all people including all LGBT people?  You would come around to his side right?

That is the difference between us and the republicans.  They will hate Obama even if he cuts taxes for millionaires, ends food stamps, and implements a gun policy that requires all Americans to own automatic weapons.  They were determined to make him a "one term president" the day he took office.

You can see it in the Sandy response.  GOP heads are exploding because Chris Christie is working with the president.  You'll notice that there are no diaries on DKos lashing out at the president for working with Christie.

• you've just made a great point.(0+ / 0-)

If transferring wealth to the top 2% creates jobs, wouldn't we be swimming in jobs right now? (-9.75 / -9.05)

[ Parent ]

• Sure(0+ / 0-)

But he's such a liar we can't trust him to do any of that.

• Their ancestors were allowed to take (2+ / 0-)
Recommended by:
white blitz, doinaheckuvanutjob

what they wanted.

To the victors belong the spoils.

Don't underestimate that.

• The Lizard People? I agree. Those Lizard People(1+ / 0-)
Recommended by:
bontemps2012

keep taking what they want and refuse for anyone else to have a sliver of their spoils.

• See, that gets into part of why I did this post.(17+ / 0-)

Math hath no anger.

But if it's all magic wizardry to you, and you don't know what Nate's doing, then when Jennifer Rabies of the Washington Post comes along and says "Nate's averaging polls, you can't do that hahahaha" you have no way of knowing how full of shit she is (except vast amounts of experience).

Pro-Obama.

[ Parent ]

• At this point, intellectual (2+ / 0-)
Recommended by:
kyril, ssgbryan

arguments have to get out of the way because the returns will be coming in soon?

• What will happen as a result of their not (2+ / 0-)
Recommended by:
kyril, doc2

accepting an Obama victory? What problems can they cause as a result? Should we be doing something in preparation to protect his re-election?

• We here on this blog, what should(0+ / 0-)

we do? Nothing.

• And they believe Reagan was a hero for (1+ / 0-)
Recommended by:
CarbonFiberBoy

getting 241 American servicemen killed in Lebanon and then doing a cut-and-run 90 days later.

But that doesn't exist if they don't say it exists?

Not up at 1984th Street.

• remember, for republicans, american history(1+ / 0-)
Recommended by:
bontemps2012
And they believe Reagan was a hero for
getting 241 American servicemen killed in Lebanon and then doing a cut-and-run 90 days later.
only began on jan. 20, 2009, nothing happened before then. bush has been disappeared, to the point that he wasn't even invited to speak at the republican convention.
• In Jersey they're running ads blaming Dems(0+ / 0-)

for "four years of increasing unemployment."

That's so bad it's not even wrong.

• What else(0+ / 0-)
They are not going to accept an Obama victory
is new.
• All this truth bursts their bubble(0+ / 0-)

People go to great lengths to avoid bursting their bubbles.  Often, violent lengths.

Conservatives need to realize that their Silent Moral Majority is neither silent, nor moral, nor a majority.

[ Parent ]

• They don't accept ANY Democratic victory(3+ / 0-)

As I mentioned somewhere else a couple days, I saw my first "Not My President" sticker during the first Clinton administration. They are zombie bullies and Democrats need to learn to twist the knife when they win, not pursue impossible bipartisan fantasies.

• So they don't accept it. Who cares?(2+ / 0-)
Recommended by:
doc2, 207wickedgood
• no one said they were a particularly (1+ / 0-)

bright bunch.

• I think Silver's response would be(3+ / 0-)
Recommended by:
Jakkalbessie, kyril, bontemps2012

... gotta base numbers on something, and the past is all we have in terms of verifiable numbers. Anything else gets you into the realm of guessing.

Pro-Obama.

[ Parent ]

• The bottom line is that(3+ / 0-)
Recommended by:
zemblan, blue aardvark, bontemps2012

If Obama wins, silver is vindicated, and if Obama loses, silver will be the goat. Nothing he can say will change that.

• It's not just Silver...(11+ / 0-)

ALL the poll aggregation sites currently have Obama winning.

If Obama loses this election, it will be an explosive blow to the science of polling.  People will be analyzing the event for years.  For that reason alone I find it unlikely the polls would be wrong this time.

------RM

• Either that, or(9+ / 0-)

It'll point to significant voter/election fraud and the effect of disenfranchisement efforts by the Republicans.

• Unfortunately not an insignificant factor(2+ / 0-)
Recommended by:
ybruti, celdd

Probabilities change when you load the dice.

Pro-Obama.

[ Parent ]

• Oh, why not both?(1+ / 0-)
Recommended by:
Wreck Smurfy

Unless someone convinces me that cell-phone-only voters are more likely to vote for Romney, though, I think there's a likelihood that the polls have a slight Romney bias.

Romney economics: Feed our seed corn to the fattest pigs and trust them to poop out jobs.

[ Parent ]

• That's basically how we got here...(0+ / 0-)
It basically says that Silver's model is based on prior experience, but that relying too much on prior experience is a mistake.
• I'm curious about cellphone only voters(0+ / 0-)

and contactees who don't cooperate with the pollsters. I would guess that cellphone only households lean Democratic. I don't have any idea about poll refusers.

• The wingers don't connect to reality when it comes(4+ / 0-)
Recommended by:
princesspat, kyril, sfbob, ssgbryan

to science in general (or for that matter, just about anything else), so why should they let something like math stop them from attacking Nate? Sounds like completely consistent behavior for them to me.

I believe that in every country the people themselves are more peaceably and liberally inclined than their governments. -- Franklin D. Roosevelt

• They slept through math class. n/t(2+ / 0-)
Recommended by:
kyril, bontemps2012

Pro-Obama.

[ Parent ]

• No. (2+ / 0-)
Recommended by:
dallasdunlap, jdld

I flunked them out, first pop quiz.

• Exactly(3+ / 0-)
Recommended by:
zemblan, CarbonFiberBoy, celdd

The polls are a reflection of the reality of the electorate. But they think of things exactly opposite - if they can unskew, cherry-pick, or otherwise manipulate the poll results, that will make their desired reality occur. It's pretty fuckin' crazy.

Nice diary, zemblan. I've used actual M&Ms to explain statistical sampling.

Progressives are defined by who they want to help; Conservatives are defined by who they hate.

[ Parent ]

• But.. but...(12+ / 0-)

...Jesus told me Nate Silver is wrong! And he's gay! Not Jesus, that Nate person! Girly man! Wrong!

Froth, bark, whine!

Fuck me, it's a leprechaun.

• Nicely done. Please re-post on Monday.(5+ / 0-)

Am I right, or am I right? - The Singing Detective

• Thanks.....(4+ / 0-)
Recommended by:
zemblan, kyril, blue aardvark, 42

I'm mathematically challenged, but I understand M & M's......and I appreciate Nate!

Love is the lasting legacy of our lives

• Thanks!(4+ / 0-)
Recommended by:
princesspat, kyril, blue aardvark, MeToo

Some of the stuff that's coming out of the GOP attacks on Silver is just so obviously wrong it's as blatant as "GM is sending your jobs to China" - but if you don't have a stats background, like, oh, say, EVERY MAJOR PUNDIT EXCEPT NATE SILVER, then you're helpless to see how you're being lied to.

Pro-Obama.

[ Parent ]

• Speaking of statistics(0+ / 0-)

Has anyone ever charted our revered pundits' predictions over the years? Take Friedman and his famous "6months until everything is okay-dokey in Iraq" predictions.  I'm really curious to see the statistical measure for how much any given pundit is talking out of his or her ass at any given time.  Then we could factor it in, and say, "Well Friedman is always  off by a factor of X when opining about foreign policy."

"YOPP!" --Horton Hears a Who

[ Parent ]

• I live in Wisconsin....(20+ / 0-)

During the Recall election of Scott Walker, most of the polls had him leading.  However, there were plenty of us here, myself included, that refused to believe it.

I could we have gatherd 1,000,000 signatures and still be losing in the polls....?  How could Walker be winning the polls after all he did to the state.  How coul dWalker be ahead in the same polls that were showing Obama was favored in Nov.  It was unfathomable.

Well, the polls were right.  In hindsight, there were enoguh voters, including Democrats, that did not believe in the Recall itself.  The exit polls showed that while the state still supported Obama enough voters felt Walker should serve out his term.  I have had a hard time grasping that logic but it was what it was.....

Anyway, while we all screamed that the polls had to be biased, they were right all along.

-5.62, -5.28

• Oh, and Silver predicted Walker would win. nt(14+ / 0-)

-5.62, -5.28

[ Parent ]

• One of the things I admire most about Nate(14+ / 0-)

is how clearly he is trying to spot places where subjectivity might be creeping into his analysis, giving him a selection bias of some sort, and then trying to eradicate those possibilities. There's nothing harder than trying to see your own flaws.

Pro-Obama.

[ Parent ]

• A.T. & T. quality control system:(1+ / 0-)
Recommended by:
Geenius at Wrok

Why everybody needs independent quality control:

1. Missed requirements.

2. Misunderstood instructions.

3. Wrong technology.

You can't fix yourself for any of the three.

Worked out decades ago. Tested by the zillions of applications.

• As I understand it.....(1+ / 0-)
Recommended by:
celdd

that argument....that Recalls should only be used under very extreme circumstances (ignoring the fact that Walker had acted extremely) was pushed heavily in robocalls and forums by Walker supporters including Rove and Koch-funded PACs.  The voters of WI, bless then, had already been through recall elections for the Legislature after a general election and I have to sympathize with them...were just bone tired of the almost continuous politicking going on and said....Enough, and voted against the recall as a result.

Free markets would be a great idea, if markets were actually free.

[ Parent ]

• I disagree. We were not 'bone tired'.(0+ / 0-)

exit polls revealed that a % of the poulation just plain disagreed with recalls - period.

More voters turned out to vote in the recall than did the initial governors election that voted Walker in.  That does not sound like a tired electorate to me.

-5.62, -5.28

[ Parent ]

• Silver isn't subjective? FALSE(14+ / 0-)

In what is an otherwise perfectly excellent summary of basic statistics you assert the following (which is self-contradictory):

But, again, the assumptions Nate makes aren't subjective, but based on analysis of previous elections, and he's gone to some trouble over the last few months to spell these assumptions out.

Nate makes many, many assumptions. Each of these are based on subjective judgment. The data themselves do not give the answer.

This is not a criticism of Nate!. His assumptions may all end up being excellent. But they are assumptions nevertheless, significantly based on subjective judgment. And a rational person could quibble with any number of them.

I guess I'm just trying to insist that social science is harder than the tone of your post suggests. I've learned a lot from Nate. I love that he uses a Bayesian approach. But his model -- as he would be the first to admit! -- isn't "just math".

If we insist that the Republicans are wingnuts because they won't look at the evidence, I think we owe it to ourselves to be as cold-eyed as we can in understanding how strong the evidence really is.

• Let me rephrase, then.(8+ / 0-)

As a human running a model, he has to at some point determine what the model's inputs are; the math books don't magically tell him whether or not to include as a variable involving whether, say, the candidate's mother's middle name includes an "i." So yes, there are subjective decisions in that sense, and it's unavoidable. You have to select your variables, and that includes judgement calls.

But the decision of which variables to include or not include shouldn't be driven by the outcome you'd personally like.

So when I mean he's not being subjective, I mean that he is choosing the variables only with an eye to whether they are useful and relevant, rather than which direction they pull his results.

One of the reasons for poll house effects is that different polling firms will use different ways to determine whether their sample adequately matches up with the general population. When it comes to people, especially relatively large groups of people, it can be difficult to answer the question: is group A like group B or not?

Pro-Obama.

[ Parent ]

• True, Ant. Assumptions aren't math, but they ...(1+ / 0-)
Recommended by:
kyril

... may be the most statistically/historically valid assumptions. My money's on Nate and his results.

Obama and strong Democratic majorities in 2012!

[ Parent ]

• We and the GOP won't have to wait much longer,(2+ / 0-)
Recommended by:
zemblan, kyril

... to see who 's closer to the result. Given his success in predicting the last two elections, my money's on Nate.

Obama and strong Democratic majorities in 2012!

• Much too wonky - makes my head spin.(4+ / 0-)
Recommended by:
zemblan, kyril, StrayCat, The Zipper

I go with the fact that there is a liberal conspiracy among pollsters biased  toward Obama and just subtract 5% from their totals. I derived the 5% from a mixture of gut feeling, common sense and listening to "Morning Joe."
Google the use of the Monte Carlo simulation to calculate Pi. Great stuff. Here's a link, though it needs a Java plugin. I am sure there are others.

• The pi example is great(1+ / 0-)
Recommended by:
kyril

... because it demonstrates that, even though at heart the calculation is hundreds of thousands of monkeys throwing dice, you really can get close to pi. As long as you have enough monkeys.

Pro-Obama.

[ Parent ]

• ... or CPU cycles.(0+ / 0-)

Einstein had a great "intuitive" solution when the math failed. Can't find it. If anyone knows what it is, I am eager to be reminded.

[ Parent ]

• Reality has a well-known liberal bias.(3+ / 0-)
Recommended by:
zemblan, kyril, MeToo

Strength and dignity are her clothing, she rejoices at the days to come; She opens her mouth with wisdom, and the law of kindness is on her tongue.

• Math too, apparently.(3+ / 0-)
Recommended by:
kyril, loggersbrat, MeToo

That's what drove me nuts about the attacks on Nate.

To paraphrase Senator Moynihan, you're entitled to your own opinion, but you're not entitled to your own plus sign.

Pro-Obama.

[ Parent ]

• A case of shoot-the-messenger.(1+ / 0-)
Recommended by:
kyril

"Life and death, dispensed on a dollar basis. How ridiculous and fatally stupid, in what is still the richest country on earth?" Exmearden

• That's really it.(3+ / 0-)
Recommended by:
kyril, celdd, burana

One of the reasons I tried to spell this out is so that you can see just how nonsensical the objections are.

Pro-Obama.

[ Parent ]

• Just like Climate Change.(2+ / 0-)
Recommended by:
kyril, burana

Don't like the message?  Besmirch the messenger and question his motives and morals.

• I calmly await the inevitable(4+ / 0-)
Recommended by:
zemblan, kyril, Clyde the Cat, StrayCat

And by that I mean, the inevitable smug pronouncements on Nov. 7 from all the pundits who will airily opine that they saw the Obama blowout coming from a mile away. And anyone who thought this was going to be a too-close-to-call nail-biter just didn't understand what the polls, in their poll-ey wisdom, were telling us poor, ignorant proles.

• tips for Patty, Maxine and LaVerne(2+ / 0-)
Recommended by:
zemblan, kyril

the rest is math to me

It's not a fake orgasm; it's a real yawn.

Pro-Obama.

[ Parent ]

• but then I wrack my brain trying to explain(1+ / 0-)
Recommended by:
zemblan

what it is that you mean

to me

It's not a fake orgasm; it's a real yawn.

[ Parent ]

• Patty, etc. (1+ / 0-)
Recommended by:
zemblan

I wonder how many of us got the reference to Patty, Maxine, and LaVerne.  I did, but I'm very old.

Maxine and LaVerne Andrews are no longer with us, but Patty lives on.  She's now 94.

I worked for the Nader presidential campaign in 2000. I'm so sorry!

[ Parent ]

• I wouldn't say I'm *very* old(0+ / 0-)

having just turned fifty. But I went to high school during the days of "Saturday Night Fever", and -- yeccch! that would turn anybody off of top twenty radio.

Pro-Obama.

[ Parent ]

• Good simplification(1+ / 0-)
Recommended by:
zemblan

And I'm recommending it because the title made me laugh.

Flee fro the prees and dwelle in sothfastnesse.

• it would eFFing hilarious if it was not(1+ / 0-)
Recommended by:
MeToo

so serious for everybody on this planet. Yes the whole planet. These people will take down everything with them when the house of cards they're building collapses.
We are today, right this minute on the "fronlines", the future is at stake here, I am not being over the top.
The attack on Math. It is time to realize that this attack on math, math which has given us the ability to do science,  which has revealed the "truth" to us in every part of our existance, math, which brought us out of the "age of superstition", math, which allowed us "see" understand and manipulate our world to our benefit is to put it in terms even these simpletons should be able to "dimly" perceive, akin to the spewing into our world of the by-product of a waste treatment plant. These flatheads deny math at the risk of another Cretaceous Tertiary extinction type event,  but this time it would be us in the cross hairs. If these christianists are so deeply wishing to see christ in person, in a so-called "rapture" event, an "event" for  which there is no evidence in the bible, then they should just pull a heaven's-gate and depart the scene, and leave those of us who embrace knowledge to continue the move FORWARD, FORWARD to a better future for ALL.
Attacking Nate is an act of desperation, which exposes these flathead as a "drag" on the existance of the human race. Let them go the way of the Neanderthals.This is "zombies at the gate".

THE TEFLON CANDIDATE-----> WILLARD MITCH ROMNEY

• Well I'm glad WE never attack people whose #'s we(3+ / 0-)
Recommended by:
Cardinal Fang, StrayCat, cardinal

don't agree with.

Oh wait, if memory serves me right, Nate got some DKOS grief for awhile there when he wasn't providing good news.

I love the look in peoples eyes when they realize, for the first time, that they have power.

• Math, like Logic(0+ / 0-)

is not calculated into political propagandist thinking. Math is for the intellectual - Propaganda is for the uneducated.

If not us ... who? If not here ... where? If not now ... when?

• and I am sure that we could use math(1+ / 0-)
Recommended by:
RUNDOWN

to prove this.

Math, like Logic
is not calculated into political propagandist thinking. Math is for the intellectual - Propaganda is for the uneducated.
to prove this.

THE TEFLON CANDIDATE-----> WILLARD MITCH ROMNEY

[ Parent ]

• Nate Silver's Math based Math Vs(1+ / 0-)
Recommended by:
doinaheckuvanutjob

Gallup's Meth-based Math !

• What I don't understand about those 18% is this:(2+ / 0-)
Recommended by:
StrayCat, doinaheckuvanutjob

If Obama is a Muslim, why hasn't he enacted Sharia law already? What is he waiting for?

Obama said knock you out.

• Maybe it's taking longer than he planned(1+ / 0-)
Recommended by:
StrayCat

to re-engrave all the signage in DC in Arabic?

Pro-Obama.

[ Parent ]

• He needs Pelosi back in charge of the House(1+ / 0-)
Recommended by:
zemblan

because seriously, all those Democratic socialist feminazi women are going to just follow whatever Obama tells them to vote for, even if it means they have to give up their drivers' licenses.

It's the House GOP that stands firmly between America and Sharia law. Just ask them.

• He has to confiscate all of our guns, first.(1+ / 0-)
Recommended by:
doinaheckuvanutjob

And everyone knows that doesn't happen until the 2nd term.

• First he's putting in fascism, which the left is (0+ / 0-)

fascist, you see, then sharia law comes later, after the communism...

I have actually read such arguments by wingers, they are in a lather of nonsense, and none of their ideas make a sliver of sense.

• I assume, therefore I am. I assume. (3+ / 0-)
Recommended by:
zemblan, pimutant, StrayCat

Excellent -- clear, succinct explanation.

I think the quibble over "assumptions" has partly to do with the mathy-sciency use of the term vs. the common use of the term.

In everyday usage, an assumption is a guess at best, and at worst an over-reliance on things one has no way of knowing.  When we say that someone is "making assumptions," it is a criticism -- implicitly we're saying they're making the wrong assumptions.

In science, particularly when running complex models, we have to make certain assumptions.  There are variables that need to be included in the calculations that we don't know precisely, or sometimes even imprecisely.  So to run the model you make an assumption: X = N, chlorophyll = phytoplankton, mortality rate = 80%, or whatever.  Or you can choose to make the assumption that a particular variable is unimportant and exclude it.

The difference is that scientists and modelers are trained and expected to be explicit about their assumptions.  When you publish results, you lay out what your assumptions were, and why you made those choices. (If you don't, you can be sure your peer reviewers will ask for it before they give a thumbs-up.)  There is usually some empirical or theoretical information that can guide the choice.  In cases where the possible choices for an assumption are very broad, there is also the option to run a model using different assumptions and compare the results, so you then have evidence of how much your assumption is influencing the outcome.  Nate routinely mentions this kind of formal or informal "sensitivity testing" (how sensitive a model is to changing a particular variable).

In short, in the mathy-sciency world, assumption is not a dirty word.  It's just something that we need to be up-front about: what our assumptions are, and why we have chosen them.  Ultimately the choice is based on what we think will give the most accurate result.  It's not the same as being subjective.

Of course someone could make deliberately off-base assumptions because they provide the most wished-for result.  That's where it becomes faith-based modeling:  I assume X because if NOT X  ... #&@%^!COGniTiVE%#DISsoNanCE&*#%!!!

Correlation may not cause causation, but I find that they often occur together. -- Mike the Krugman commenter

• this is one "approach",(0+ / 0-)

which is applied in the field of medical technology, which was the path on which I chose to use my G.I. bill education benefit.

THE TEFLON CANDIDATE-----> WILLARD MITCH ROMNEY

[ Parent ]

• The problem is that the polls capture(1+ / 0-)
Recommended by:
StrayCat

how people are intending to vote, and then as in Florida 2000, tens of thousands thought they had voted for Gore and either voted wrongly (butterfly and caterpillar ballots) or their ballots were thrown out for technical reasons.  That must be why Nate now had a less than 80% probability that Obama will win - as he says, anything can happen.

The spirit of liberty is the spirit which is not too sure that it is right. -- Judge Learned Hand, May 21, 1944

• Thanks, good diary.(0+ / 0-)

The way I see it:

Mr. Silver's job depends on his accuracy.

So the question is, does he want to keep his job?

• Something else you should mention(7+ / 0-)

if you repost on Monday (an idea I heartily endorse):

There's a common misconception that the distribution of outcomes within the margin of error is flat.  Consequently, given a poll that has R 49 - O 44 with a 3% MOE, you'll often see posters say something like, "Well, that's within the MOE and it could actually be Obama +1!".

That is, of course, extremely unlikely, as the distribution of outcomes follows a normal curve.  A result where Romney's support was overstated by 3% and Obama's was understated by 3% would only occur 1 in 40 times.  Yes, it's possible, but it's probably not the case.

Unfortunately, though, it's very common for people to see these polls in terms of binary outcomes.  They assume that all outcomes within the MOE are equally possible, and also assume that all outcomes outside the MOE are equally impossible.

To believe that markets determine value is to believe that milk comes from plastic bottles. Bromley (1985)

• This is why more elementary math classes(3+ / 0-)
Recommended by:
EclecticCrafter, birdboy2000, bara

should be devoted to rolling D&D dice.

Pro-Obama.

[ Parent ]

• I gave extra credit for D&D when I (2+ / 0-)
Recommended by:
zemblan, StrayCat

Okay, not really, but I did accept a project based on rolling D&D dice.  And I played D&D with some of the kids & parents.  But I would have given credit if I didn't think it would have gotten me fired.

• Oh, wow. I might have had MOE wrong(1+ / 0-)
Recommended by:
sneakers563

for all these years.

In your example above of R 49- O 44 there is a 5 point edge to R.  I had always thought that a MOE of say, 3, meant that the 5 point edge to R, could possibly go up to an 8 point edge or down to a 2 point edge for R.  And since the lead was 5 points and the MOE was 3, that R's lead was outside the MOE.  But you are apparently telling me that a 5 point lead is within the MOE because you could possibly subtract the 3 points from the leader AND add 3 points to the trailer.

I trust the knowledge base on this thread and so can you all set me straight?  I'm a big boy and can handle being wrong in the past.  I just don't want to be wrong going forward.

Thanks.

We cannot solve our problems with the same thinking we used when we created them. Albert Einstein

[ Parent ]

• That's correct(0+ / 0-)

The MOE applies to both of the individual statistics, not the difference between them.

However, things aren't so dire as they seem (or as great, depending on your perspective).  Because there are two probablilities involved, you multiply the respective MOEs, so the chance that both statistics are wildly off is rather low.

For instance, I made a mistake in my original message.  The MOE is the certainty that a particular number is within that range, assuming a random sample.  However, because the actual result can be either above or below, the remaining 5% chance is split between the two.  There's only a 2.5% chance the actual number is greater that the MOE, and a 2.5% chance the actual number is less than the MOE.  Therefore, with a 3% MOE, the chance that one number is 3% or more low and the other is 3% or more high is 2.5% * 2.5% = .06%.  That's 6 cases in 10,000!  It would be extremely foolish to call a 5 point lead "tied" in that case, despite it being "within the margin of error".

In that sense, if you're looking for reassurance, your original approach was probably better, even if it's not 100% correct.  Realistically, assuming one of the numbers is correct probably gives you a better sense of whether you should be worried.

To believe that markets determine value is to believe that milk comes from plastic bottles. Bromley (1985)

[ Parent ]

• Yes! That's one of my(0+ / 0-)

biggest pet peeves in news coverage of polls. I second the call to add it to the diary in the event of a repost.

Grew a mustache and a mullet / Got a job at Chick-Fil-A

[ Parent ]

• The part that seems to cause confusion..(2+ / 0-)
Recommended by:
StrayCat, Leap Year

..relates to adjusting the responses based on demographics (which all pollsters do) and party ID (which only a few do).

If PPP calls 100 people, and 40 men respond, but only 22 women do, they are going to weight the responses from females much higher to get the proportion back to accepted US Census numbers for the region.  They generally do this for gender, age groups, race, and sometime education levels.. things that are known, and don't change very often.

However, if PPP calls 100 people, and 40 Dems respond, but only 22 Republicans do.. those are the numbers they are going with.  Why the difference?  Because party ID changes.  There is no way to accurately weight responses without putting the pollster's preconceived notion of political leaning in the way!

This is how Rasmussen.. one of the few pollsters to weight by a party ID percentage that exists primarily in their own spreadsheets.. can adjust the results as they choose in the months leading up to the election, and then conveniently shuffle them back to reality right before election day.

When extra-terrestrial beings make their first appearance on our planet, and ask for representatives of our species to best exemplify humanity, I'm sending a nurse, a librarian, and a firefighter.

• just a quibble...(0+ / 0-)

I think we can pretty much count on PPP doing some deletion and weighting to every sample, regardless of how many Democrats and Republicans there are in the unweighted sample.

It's the weighted composition of Democrats and Republicans that they will not further tinker with: they will not weight to party ID.

That distinction may be worth making because sometimes people seem to think that unweighted Ns tell them the composition of weighted samples. They don't. But sometimes they give some insight into how much a sample is being stretched.

For instance, if (silly hypothetical) a pollster manages to interview 20 people under age 45 in a thousand-person sample, that doesn't mean that their weighted sample will be 98% 45-plus. But it does mean that their estimates could be pretty awful, because 20 interviews aren't likely to represent the under-45 set very well.

Election protection: there's an app for that! -- and a toll-free hotline: 866-OUR-VOTE
Better Know Your Voting System with the Verifier!

[ Parent ]

• Was that a quibble, or just an echo?(0+ / 0-)

I can't find any disagreement.. in fact, it's almost a mirrored copy of my comment.

When extra-terrestrial beings make their first appearance on our planet, and ask for representatives of our species to best exemplify humanity, I'm sending a nurse, a librarian, and a firefighter.

[ Parent ]

• here....(0+ / 0-)
However, if PPP calls 100 people, and 40 Dems respond, but only 22 Republicans do.. those are the numbers they are going with.
Probably not, because they'll upweight the young respondents -- so I suppose they will probably end up with even more Democrats. (OK, so the numbers are weird!)

I assume I'm saying what you meant, but not everybody "gets" this.

Election protection: there's an app for that! -- and a toll-free hotline: 866-OUR-VOTE
Better Know Your Voting System with the Verifier!

[ Parent ]

• They don't weight for Dem/Rep.(0+ / 0-)

Those are the numbers they are going with.

I didn't say anything about the other characteristics of those respondents, but the Dem/Rep mix of their polling will be 40 and 22.

And I don't see how you can argue with that statement, as you repeated it basically word for word.

When extra-terrestrial beings make their first appearance on our planet, and ask for representatives of our species to best exemplify humanity, I'm sending a nurse, a librarian, and a firefighter.

[ Parent ]

• Terrific Diary(4+ / 0-)

Just the right touch of geek mixed with the reality of the cosmic shit storm that people on the far right seem to thrive in.

• I'm curious, though(1+ / 0-)
Recommended by:
bontemps2012

about what happened 4 years ago in the Dem primary in New Hampshire?, I believe, when the polls showed Obama ahead of Hilary by 8-10 points the day before the vote.

• Primaries are harder to forecast(0+ / 0-)

(opinions are much more fluid). Additionally there was not much time after Iowa and before NH, to see if the poll results would be sustained.

• From your friendly local...(1+ / 0-)
Recommended by:
zemblan

statistician, I say well done zemblan.

I would never belong to a club that would have me as a member--Groucho Marx.

• Outstanding diary - clear and understandable(2+ / 0-)

I though I knew Nate's methodology pretty well but you illuminated a lot more for me.  Thanks.

Regardless, the wingers will cover their ears, close their eyes and scream "la-la-la-la-la."

• Mmm telling?(3+ / 0-)

Nate Silver will be on The Daily Show (again!) the night after the election. Bragging perhaps?

• Was he on the Daily Show already?(1+ / 0-)

Damn I missed it! I just finally watched the POTUS episode last night.

The man who moves a mountain begins by moving away small stones. -Confucius

[ Parent ]

• Another thing to consider...(1+ / 0-)
Recommended by:
princesspat

In statistics, a general rule is garbage in... garbage out.  Let's say a given poll is biased... not specifically calling out Rassmussen or anything but if the shoe fits...

Let's say I am conducting a poll, I have to make assumptions about distribution and GOTV in order to calulate where to get samples from.  These assumptions are critical to my poll because it removes randomness and inserts bias.

Lets use your m&m analogy.  Let's say I reach in a bag that I know contains 1000 m&ms and I take out 50 m&ms.  If they are 25% red, I can estimate that there are 25% red m&ms in the entire population (plus or minus a percentage that is based on charts that I do not have available right now).  However, there are assumptions built into this estimate that may not occur to some people like, were the bag of m&ms shaken up so that they were randomly distributed?  Are all of the m&ms the same size and weight? Did I have any unintentionally bias in the manner in which I counted them (exluding broken m&ms, counting two broken ones as one, etc...)?  The same thing goes with polling.

If I only use electronic polling, would there be a greater liklihood for a teenager or an elderly person to respond?  Would someone of means be more likely to spend time responding than a poor person?  Who would be more likely to lie?  and on and on and on.  If the root data is wrong, any data derived from the data will also be wrong.  Nate Silver does a good job of trying to account for bias and uses a large sampling for a bigger average but garbage in, garbage out still applies.  The 95% confidence number assumes good, unbiased data.   I can get a 95% confidence level from crap data just like I can from good data.   The only way to know for sure is to do additional studies regarding the polling methodology which Nate may have done.

The point I am trying to make is that polling is an inexact science for many reasons.  In reality, it is all about who gets the most people to come out to the polls and vote.  Talk to your friends, talk to your neighbors, offer them a ride to the polls, increase your vote by a multiplier by taking additional people with you.  GOTV is the most reliable way to win an election.  This comes from a guy who crunches numbers for a living and I love statistics.

"Perhaps the sentiments contained in the following pages, are not YET sufficiently fashionable to procure them general favour..."

• Set the wayback machine two weeks(5+ / 0-)

Back when Nate's model was showing the race nearing 50-50 territory. I can remember a few Democrats starting to pull out the knives to attack him.

Now suddenly Nate is every Democrats friend.

I say this not to criticize Dems but to point out that Nate's model is not partisan. When it contained bad news for Obama, he published it. When it contained good news for Obama, he published it.

May you live in Interesting Times

• You know, I stuck with Nate(2+ / 0-)
Recommended by:
cardinal, SneakySnu

during those weeks, because the math is the math. But I think it aged me a year.

Pro-Obama.

[ Parent ]

• Indeed.(0+ / 0-)

The good news is, we've got it right now. And I'm happy to say that I vigorously defended him here when folks were attacking him for not producing the desired outcome.

Grew a mustache and a mullet / Got a job at Chick-Fil-A

[ Parent ]

• What Nate doesn't seem to correct for...(1+ / 0-)
Recommended by:
Leap Year

... and what I worry most about is the effect of the voter suppression / intimidation tactics. I don't think you can account for what we're seeing this year based on historical data. Looking at how far they've gone to stop people from even getting a ballot in truly makes me wonder what other tricks may come out of the bag come count time. Especially with out-of-band early / absentee votes.

Maybe I just voted in FL one too many times, but I have no faith in the gatekeepers anymore.

• Silver's models(0+ / 0-)

do take voter suppression measures into account. At least in his "state fundamentals" calculations.

• The absolute numbers are not as important...(1+ / 0-)
Recommended by:
Leap Year

...as the trendlines.

Nate's model makes laughable the Mittmentum talking point everyone had last week. If anything, it shows momentum in Obama's favor.

May you live in Interesting Times

• What do you mean, "last week"?(0+ / 0-)

Jennifer Rabies is still trying to push it at the Washington Post.

Pro-Obama.

[ Parent ]

• At the place where...(0+ / 0-)

...you said calculus was required, I don't believe it was.  Calculus doesn't get involved from time to time, but not with confidence intervals, margin or error and sample size.

• Well, what I was thinking(1+ / 0-)
Recommended by:
rserven

was expressing the actual formulas for calculating the various values that you generally use lookup tables for if your calculator (or stats software) doesn't already have it.

Using the various tables doesn't require calc, but calculating the tables themselves in the first place does.

Maybe I just put that badly.

Pro-Obama.

[ Parent ]

• Well done and highly informative. You have (2+ / 0-)
Recommended by:
StrayCat, zemblan

answered a number of questions clearly that my wife has asked me, but you made it much more lucid than I.

Also the discussion was equally illuminating.

good job.

"There is nothing - absolutely nothing - half so much worth doing as simply messing about in boats ..." - Kenneth Grahame -

• Nate will always be Poblano to me n/t(5+ / 0-)

"Life is a bitch, and then you die. And then you come back." Old Buddhist proverb

• Your poll question would be more fun if.....(2+ / 0-)
Recommended by:
zemblan, MeToo

Your poll question is funny, but it would be much more fun and interesting if you asked the same question but with 5 possible answers:

The probability of the first poll item winning the poll:

a) 0%
b) 20%
c) 50%
d) 80%
e) 100%

Think about that one for a second and it becomes an interesting sociological experiment. The distribution of responses to this poll could be very amusing.

• Mental masturbation.(0+ / 0-)

That's what Nate and Kos and the crew are struggling against.

The righties are lost in left-brain non-rational horn-polishing.

Their perfect candidates were Reagan and Scott Brown.

Reagan was a genius, which he disguised with the skills of a Hollywood leading man. Not many fooled Clark Clifford. He did it effortlessly.

"Amiable dunce."

Brown is an amiable dunce, at his best.

Beating these SOBs requires winning ugly. Winning lots of votes ugly, anyway.

• There's a method to the Right's madness(1+ / 0-)
Recommended by:
Leap Year

I believe this is more than sour grapes and trying to get their people to the polls -- both are factors, but I think there may be another issue here, namely...

The Right has a pretty good idea that Romney is going to lose.  When he does, they want to make the case that the election was stolen.  That's harder to do if the best statisticians in the business are telling them a Romney loss was likely, and the election results are therefore likely valid.

• I don't think you need to suppose that.(1+ / 0-)

I think the real problem is more visceral than that: some geek is telling them that their reality distortion field has got it all wrong, and so they blame the geek, not the field.

Pro-Obama.

[ Parent ]

• And not even necessarily(0+ / 0-)

make a unified case that the election was stolen (though I'm sure some of the radio talking heads will), but at least to ensure that the bitterness and outrage of the far right continues. Rush needs listeners and FOX needs viewers to be angry for the next 4 years!

• Listened to Hannity (Briefly) Today(3+ / 0-)
Recommended by:
42, VTGenie, General Hubbub

He'd abandoned the Benghazi "Coverup" libel for the moment to bring in a conservative media windbag from Ohio who declared the state is solidly for Romney. They were also very careful to talk about 2 polls that show Romney with a lead in Ohio and Obama in trouble. Then they started gassing about various states and the chances there.

The whole point isn't to do honest reporting (Duh!) but rather to keep feeding the narrative that Obama is going to lose and Romney is going to win. They're trying to keep the faithful in line, and as authoritarians do, they're trying to control what they see and hear.

This could be a real problem if Romney doesn't win. They are prepping millions of people to believe the election process is corrupt, that the election was stolen, and they're priming them to act in the worst way.

Anyone who thinks an Obama win is going to send a message to Republicans to chill their jets a little and start thinking about working together with Democrats for real has their head up their butt. It's far more likely they're only going to adopt even more of a bunker mentality and behave badly.

"No special skill, no standard attitude, no technology, and no organization - no matter how valuable - can safely replace thought itself."

• xaxnar - This could backfire.(2+ / 0-)
Recommended by:
xaxnar, prairiegirl

If the T-Pub base believes that the electoral process is rigged, they might give up on voting.

• Someone make me feel better!(0+ / 0-)

I have some political savvy conservative friends and I'm confused as to how they honestly believe in a Romney landslide.  I'm not one to laugh and call them crazy because they're not.

So I compared WI and OH exit polls from 2008 (from CNN) to the most recent polls (that showed all their data) and found that most of them are pretty close to the 2008 exit polling party affiliation.  If anything both R's and D's are oversampled and I's are undersampled.

So these three points:
-I saw that in most I's were up for Romney by a decent amount.

-I don't think turnout is going to be as good for Obama as it was in 2008.  Not many are predicting as big of a win as we got against McCain

-Right now the state polls are pretty close.

Doesn't that justify me being scared of a loss?

• Perhaps a glance(1+ / 0-)
Recommended by:
123man

at NMMatt's comment above?

Something pithy TBA

[ Parent ]

• Monte Carlo(1+ / 0-)
Recommended by:
zemblan

In 1873, British engineer, Joseph Jagger, had six clerks covertly keep records of roulette wheel spin results at the Beaux-Arts Casino in Monte Carlo, Monaco. Based on the information from tens of thousands of spins, he could determine the wheel's mechanical shortcomings or its bias. In particular, he determined that one of the six wheels had a significant bias that led to the numbers 7, 8, 9, 17, 18, 19, 22, 28, and 29 to result more often than other numbers. The casino, realizing that they had been mathematically compromised, tried a variety of tricks, swapping wheels, for instance. Jagger was their equal and solved their puzzles.

Although he never did "break the bank," he was known as The Man Who Broke The Bank at Monte Carlo.

Ulam named his theory on the number of simulations leading to a statistical truth, or informally, the Monte Carlo method.

Like card counters, people who attempt to determine a roulette wheel's bias quickly become persona non-grata in today's casinos.

Recommended by:
bara

Pro-Obama.

[ Parent ]

• With a career as a math-based person let me say(4+ / 0-)

I approve this message.

I believe that you have honestly translated what Nate does into English.  Although I regularly work in statistics, there is an understanding that statisticians inhabit a mafia of understanding.  I would never rewrite a statistician's interpretation of a result, for example, without the statistician's explicit approval.  With that understanding, however, everything in this article seems to me to be accurate and entirely consistent with my understanding of Nate's model, including correlation, covariance and all the other sordid details.

You must be the change you want to see in the world. - Mahatma Gandhi

• Thanks!(0+ / 0-)

I thought it might be a good chance for some demystification.

Although it's also worth pointing out that there are additional wheels within wheels dealing with how Nate handles sparse polling data early in the election cycle by extrapolating other indicators -- but sparse polling sure isn't a problem at the moment.

Pro-Obama.

[ Parent ]

• T+R(0+ / 0-)

An actual analysis of how polling work is done? Say it ain't so!

- Polling is supposed to be a black art that involves drawing dark circles on the ground in blood

-- Or chocolate

--- Mmmm

(-5.50,-6.67): Left Libertarian
Leadership doesn't mean taking a straw poll and then just throwing up your hands. -Jyrinx

Pro-Obama.

[ Parent ]

• One of the things you didn't mention here, (0+ / 0-)

and it is why every day I look at Nate's column first.  He was the ONLY pollster who nailed the election last presidential election.  He was off by some tiny amount, and nailed all the races, both presidential and down ballot.  He has figured out a methodology that factors in even such things as the Bloomberg endorsement and Hurricane Sandy.  They are all in there.  The reason the Fox people have been running around with their shorts in a wad, is because he is showing a consistent Obama poll upward, and an amazingly steady polling in the killer states.  There are little things, like when Ann Romney went and yelled at that little group of ladies in Iowa, and it went viral in Iowa, and even here in Minnesota, will that affect the polls?  YOU BETCHA.  People from the midwest don't like people to lie to them, something that Romney doesn't understand whatsoever, and neither do they like discourteousness.  Ann's "performance" just because she was asked a question she didn't want to answer, I think lost her husband Iowa.  All those women went home, called their sisters and husbands and they called their neighbors and pretty soon it was all over Iowa, and THEY DO NOT LIKE HER ONE LITTLE BIT.  All of it matters.  Most of all, character matters, and not showing your tax returns, not telling us that all your money is in tax shelters in the Caymans, all matter.  No matter the numbers, it is the GOP nominee that is flawed, not the numbers.

"My Momma always taught me to play by the rules, and if you don't that's called cheatin'." - Donna Brazile

• Thanks!(1+ / 0-)
Recommended by:
zemblan

This was very clear, and very helpful to us Humanities people who ran from math like Republicans from reality.

• At first I thought it read "Meth-Based Math"(1+ / 0-)
Recommended by:
zemblan

but that's something else entirely.

• For Meth-Based Math(0+ / 0-)

you'll have to try some other party.

Pro-Obama.

[ Parent ]

• Oy, probstats.(0+ / 0-)

I enjoyed them just a teeny bit more than Number Theory.  Which I despised.

Says the person with a BA in Mathematics.

"So, please stay where you are. Don't move and don't panic. Don't take off your shoes! Jobs is on the way."

• I like number theory better, actually.(1+ / 0-)
Recommended by:

Just less call for it in the real world.

Pro-Obama.

[ Parent ]

• Great post(0+ / 0-)

Awesome post!  Hopefully it will help calm some of the worriers.

One thing I wish you had added here was analysis of WHY Unskewedpolls.com is so very, very wrong.
I'll give the stage to Mark Blumenthal

1. Party identification is an attitude, not a demographic characteristic. "Generally speaking, do you consider yourself a Republican, a Democrat, an independent or what" is basically the question most pollsters ask. People can change their minds about which party they consider themselves closer to. They cannot change their age, gender or race (at least not easily).
4. Claims that media polls "assume" a specific partisan or demographic composition of the electorate are mostly false. The pollsters behind most of the national media surveys, including those who conduct the CBS/New York Times/Quinnipiac, NBC/Wall Street Journal/Marist and Washington Post polls, all use the same general approach: They do not directly set the partisan or demographic composition of their likely voter samples. They first sample adults in each state, weighting the demographics of the full adult sample (for characteristics such as gender, age, race and education) to match U.S. Census estimates for the full population. They then select "likely voters" based on questions answered by the respondents, without making any further adjustments to the sample's demographics or partisanship.

Minority rights should never be subject to majority vote.

• So why have an election?(0+ / 0-)

I find these Nate Silver posts to be dangerous  We need more not less to get out and vote.  Declarations that math says Obama can't lose, is just reckless.

• You're missing the whole point.(0+ / 0-)

The math doesn't say Obama can't lose at all. Go to Nate's site and he'll tell you Obama has a 20% chance of losing. That's a far cry from 0%. I'm not about to play Russian roulette with a bullet in one of the gun's five chambers.

"Sometimes a scream is better than a thesis." -Ralph Waldo Emerson "YEAAAAAAARGH!" -Howard Dean

[ Parent ]

• I'm very careful to distinguish between(0+ / 0-)

probabilities and predictions.

Nate is about probabilities first; the probabilities imply predictions.

But if we don't get out and vote, then Romney wins.

I don't see Silver's message as "stay home and don't worry"; I see it as "Ignore the GOP fogbank about Fauxromnentum."

Pro-Obama.

[ Parent ]

• There is an easier and less open to criticism(1+ / 0-)
Recommended by:
zemblan

way, that of Sam Wang (Princeton Election Consortium). He only takes state polls, does not correct any 'house effects' and uses the median instead of the mean value (in effect devaluing outliers - an outlier can push the mean seriously, but will influence the median very little).

And his numbers are currently 96% probability for Obama reelection.

But let us not take anything for granted, so whoever can, please GOTV.

He who can make you believe absurdities, can make you commit atrocities.

• zemblan, thank you for this very(0+ / 0-)

InterestIng summary. Question  that I hope you might have a moment to answer: where in the calculation , if anywhere, does he account for past performance measured against reality for each pollster? Is that in the bias?

Thanks.
Pg

• As I understand it(1+ / 0-)
Recommended by:
prairiegirl

When he calculates the current state of, say, Wisconsin, he doesn't just average all the corrected polls, but he weights them so that more recent polls count for more and older polls fade away.

I think that weighting factor also includes greater weight for human-calls versus robo-calls, and I also think he also gives less weight to pollsters he's found were utterly wacky.

I'm not completely sure, but I think that's the idea.

Maybe someone with a better memory can remember the polling outfit he unveiled as fraudulent earlier this year?

Pro-Obama.

[ Parent ]

• Trust Nate(0+ / 0-)

... but work like hell to get out the vote.

If Nate calls the election correctly again, we'll all hail him as the prophet.  But in 2014 or 2016 if his predictions are against Dems, then we can't start namecalling and worrying that his methods are incorrect.  We'll have to work like even more hell and get out the vote.

• Not to be a noodge, but this is a question sorta..(0+ / 0-)

Isn't the "4000 holes in Blackburn Lancashire " line addressing the folks who fill the Albert Hall, meaning the hall holds 4000? Always sounded to me like Lennon calling them assholes. I'm mean though, and snarky.

Have you pampered your born-again? If you have, try.... www.getequal.org

• "A Day in the Life"(0+ / 0-)

was inspired in part by one day's newspaper, including an article about a road study in Blackburn that found four thousand potholes. Lennon added the rhyme about Albert Hall.

Pro-Obama.

[ Parent ]

• I'm still calling satire.(0+ / 0-)

It has to be.

Have you pampered your born-again? If you have, try.... www.getequal.org

[ Parent ]

• As has been pointed out before, the Right doesn't(0+ / 0-)

believe in math!

Old Hippies Never Give Up!