This! This! This all day long. Without comment

[LaydeeBug]

Link to tweet

[oasis]

1. Putin with help from his friends in Trump Tower.

We're finding out more and more daily.

[dalton99a]

33. +1

[ProudLib72]

2. But math is science and science is sorcery.

It was God's will that tRump became president.

I can't believe I am saying this, but I imagine it to be exactly what the deplorables believe.

[mastermetaphysics]

28. Lord...

When you Eat a Meal... You are Tasting Yourself...

http://mastermetaphysics.blogspot.ca

[BlancheSplanchnik]

50. I didn't need that image.

😆

[malaise]

3. Stolen

We know that

[TheDebbieDee]

4. I believe the vote counts were flipped in some of the states

that use wifi/internet to aggregate and tabulate votes... The FBI may be investigating that, too before this Trump/Russia thing is all over...

And guess what? We don't know WHO might try to meddle with the vote totals in 2018...

[INdemo]

79. I believe the votes were flipped in 2016 and 2004

For this 2016 election, there are just too many variables and proof that Donald Trump did not win.

so the point I'm want to make is we know and there are those that can prove our suspicion but why doesnt Democrats or why didnt the Democrats scream from the roof tops that this election was STOLEN

[SCantiGOP]

5. A math teacher

Sure can't argue that's not an authoritative source.
Posts like this make the entire forum look like a conspiracy site.

[NBachers]

7. Not to me they don't

[mountain grammy]

9. Don't agree

Stolen election.

[iluvtennis]

70. +++ agree

[Lucky Luciano]

80. Could be stolen, but this isn't the evidence.

[american_ideals]

10. Yes-wrong. Statistician here. Problem is polls may have been flawed.

The probability calculations are invalid unless the poll numbers were accurate reflections of the way people would vote.
And we know one trick of the GOP was to tell their voters to lie to pollsters (to lull HRC voters into a false sense of security so they'd be ok staying home.)

Under that assumption this calculation is wrong.



p.s. I believe the data-driven GOP (plus potentially russian) voter suppression campaign, plus Chaffetz's illegal leak of Comey's letter, plus cross-check and minority voter suppression all led to a stolen election. I just don't believe these data support the conclusion the vote was irregular.

[Lucky Luciano]

13. Not to mention .16% is probably based on the 5 states being independent events.

Last edited Sun May 28, 2017, 06:36 AM - Edit history (1)

They are highly correlated events of course!

[american_ideals]

39. Yup

Although the key is whether fluctuations around the mean are independent.
(Here they probably are not, e.g. because cross-state news events could affect fluctuations.)

[Lucky Luciano]

78. Hmmm

<nerd>
I think they were just straight up multiplying probabilities of a dump win in each state. This is fine if the overall events are independent.

Correct me if I am wrong below - I use some stats, but I am not a statistician:

If we assume that the polls are from unbiased samples, then the probability of dump winning one state can be calculated based on the assumption that the proportion of pro-dumpers in the sample is asymptotically normally distributed (central limit theorem). The joint distribution of the five states will be independent if the covariance matrix is diagonal since it will be normal and the second moment is sufficient to establish independence. Was it this covariance matrix, which of course implicitly handles fluctuations about the mean the basis for your comments?

I would probably say (though I don't have the data to support it) that the first principal component probably explains 90% or more of the variance in the joint distribution. The second component probably gets 99%. These really are highly correlated states. Maybe NC/FL ate a bit less correlated and would cause us to need the second principal component.
</nerd>

[american_ideals]

105. I was referring to bias in the means but such covariance could cause problems too.

also - hard to calculate principal components because you need multiple replicates; here we only have one election.

[L. Coyote]

81. They are independent polls, all robust with determineed margins of error.

Not unlike doing five separate polls about the outcome of a sporting event at five different exits. What are the chances people would get the score wrong by so much at all five exits and get it wrong by the same amount in the same direction?

[Lucky Luciano]

84. Not what I mean by independent events. Think of it like this...

If I ask you the odds of Hillary taking Texas, you would probably say about 5%. If I tell you she took Oklahoma and then I ask you the odds of her taking Texas, you might say 90%! Taking Texas and taking Oklahoma are NOT independent events.

On the other hand, if you flip a fair coin and it gives heads 10 times in a row, would you say the odds of an eleventh head on the next flip is anything but 50%? No - because it is a fair coin and past flips mean nothing for the next flip.

[L. Coyote]

86. I'm saying is you can do the coin flip test five times and the odds are you will get 50% all five.

What the OP says is that the odds are only P = 0.0016 that the five tests could all be off so much and all five in the same direction.

The same rules apply to coin flipping as polling. You don't have two separate statistics sciences for coins and other events.

[Lucky Luciano]

91. .

[progree]

92. Yup, its hopeless. Some people have no understanding that the factors that make a poll wrong

in one state might -- just maybe might -- apply to other similar states too.

And the "determined margins of error" referred to in #81 is bullshit too -- those margins of sampling error percentages assume a perfectly random sample, and that the only limitation of the poll is that the sample size is small. Those error percentages do not include the many other kinds of errors like people not answering their phones to unknown numbers (more so in some demgraphics than others), less civically engaged people (e.g. Trumpistas and Jeepers) less likely to accept a request for a polling inteview or to complete one, the cell phone vs. landline problems, and the list goes on. And reading about all the ways they adjust the polls to get the right mix of demographics is like watching sausage being made. (I had to laugh when somebody earlier in the thread said they have ways of adjusting for people lying.)

[L. Coyote]

93. Your level of misunderstand is remarkable.

Inferential statistics has margins of error to account for the factors which make sampling inaccurate. And sampling involves ascertaining that the sample is representative of the group. How many graduate level statistics courses have you taken? How many college math classes have you taught? I'm guessing zero and zero. Have you published statistical studies? Again zero, right.

[BzaDem]

97. That is a quite remarkable misunderstanding of basic high school (let alone graduate) statistics.

"Inferential statistics has margins of error to account for the factors which make sampling inaccurate. And sampling involves ascertaining that the sample is representative of the group."

The reported margin of error in a poll takes into account sampling error. It does not take into account modeling error.

The correct population for a political poll is the universe of actual voters on election day. A pre-election poll is conducted before the election (by definition). Because of this, the universe of actual voters is unknowable with certainty. Whenever you are asking human beings to make a prediction about whether or not they will take a certain action (such as voting), many responses will be inaccurate. And nothing from statistics says (or possibly could say) that such inaccuracy in predicting the future is uncorrelated from poll to poll. (In fact, we know that when we ask people who they DID vote for in the past, without any prediction, a small percentage will consistently claim they voted for the winner despite not having voted for the winner. This has been shown over and over.)

Furthermore, even if we discount the problems inherent in asking humans to predict their own future actions, getting an actual representative sample is extremely difficult when 90% of people you contact don't answer their phones. Pollsters use weighting techniques to try to massage their extremely non-representative data into a representative sample, but such weighting necessarily involves making assumptions about the electorate's composition that could be wrong. And the idea that such modeling error is independent from poll to poll is ludicrous. (It isn't even taken into account when calculating the margin of error, because there isn't even a quantifiable way to take it into account prior to the election.)

In this election, the polls slightly under-represented non-college educated whites, even after correcting for other factors. This naturally had a disproportionate effect in states with large numbers of non-college-educated whites, which unfortunately for us included critical swing states.

(Similarly, in 2012, the average of national polls had Obama at +0.7%. He won by 3.7%. This overperformance was even larger than Trump's overperformance. The proability of such a disparity due to sampling error alone, after combining all pre-election polls, is very low. But such error is utterly commonplace, and that is because of error inherent in modeling a future electorate.)

[progree]

98. You need to read up on polling, and on margins of sampling error

Last edited Mon May 29, 2017, 09:47 AM - Edit history (2)

and what makes you think he is a college math teacher? Anyone who can't convert a probability of 0.0016 into a percentage very unlikely has taught above the elementary level. I doubt he or she is even a math teacher, but it's possible; I've had some dumb ones.

>> Inferential statistics has margins of error to account for the factors which make sampling inaccurate. And sampling involves ascertaining that the sample is representative of the group. <<

Well they do try. BzaDem explains it well. Please read and absorb. You can also Google polling margin of error. I've read a number of them in the past. Here's just one example:


-----------------Begin excerpt from Pew Research --------------------------------------------------
It is also important to bear in mind that the sampling variability described by the margin of error is only one of many possible sources of error that can affect survey estimates. Different survey firms use different procedures or question wording that can affect the results. Certain kinds of respondents may be less likely to be sampled or respond to some surveys (for instance, people without internet access cannot take online surveys). Respondents might not be candid about controversial opinions when talking to an interviewer on the phone, or might answer in ways that present themselves in a favorable light (such as claiming to be registered to vote when they are not).

For election surveys in particular, estimates that look at “likely voters” rely on models and predictions about who will turn out to vote that may also introduce error. Unlike sampling error, which can be calculated, these other sorts of error are much more difficult to quantify and are rarely reported. But they are present nonetheless, and polling consumers should keep them in mind when interpreting survey results.

More: http://www.pewresearch.org/fact-tank/2016/09/08/understanding-the-margin-of-error-in-election-polls/

----------------End excerpt------------------------------------------------------------

[LeftInTX]

15. Mathematician here. I agree

I haven't been active in math for a long time, but Nate Cohn gave Trump a 30% chance of winning.

[Squinch]

53. I wonder what the percent chance was that Trump would cheat by messing with vote totals.

[LaydeeBug]

30. You mean, you castigated this OP on an ASSUMPTION you made?

wow

[american_ideals]

38. No. Must rule out simpler explanations

Unless the OP rules out alternatives, the OP cannot make that conclusion.

That's how science works.

[LaydeeBug]

41. and in this instance, the 'simpler explanation' is supposition.

you don't get to claim the mathematical high ground while you're standing on a heap of bullshit.

[american_ideals]

52. Distortion of the scientific method.

A claim is only valid if alternative explanations are ruled out. If a claim is invalid, nothing can be concluded.

Here, an alternative explanation has not been ruled out. Therefore nothing can be concluded.

[LaydeeBug]

55. ...is exactly what that's doing.

[L. Coyote]

82. No. Statistics isn't about ruling out other explanations. It is mathematics.

You design a very robust poll with lots of respondents so you have a small margin of error. This happened over and over again in the battleground states and with exit polls.

What the OP says it that the probability that all five were wrong by the amount they are wrong by is a certain probability. That's all math.

The only supposition seems to be Putin. But, who needs Putin when the Republicans already steal elections?

[american_ideals]

103. Sorry, you can only reject the null if your model is correct

That is how statistics works. You have to make assumptions to construct your null hypothesis, then you test against the null, then you reject the null if evidence is sufficient.

>What the OP says it that the probability that all five were wrong by the amount they are wrong by is a certain probability. That's all math.

The p in the original post is computed assuming a null in which the poll means are correct. The math only holds if the poll means are correct; we believe they were not because republicans were told to lie to pollsters.

p.s. love your posts in general

[tomp]

44. not a statistician...

...but...my logic begins with the fact that worldwide, exit polls are used to measure the validity of the vote count, and these did not match up in the u.s. in 2016 in some critical cases.

are you by any chance saying this is no longer a valid means of determining the fairness of elections, or that 2016 is a special case for some reason?

[american_ideals]

51. Neither. Just OP's conclusion isn't supported.

OP is using poll data and vote outcomes to determine if there was voting irregularity.

What I am saying is if poll data was distorted, the probability of voting irregularity calculated above is wrong.


There may have been voting irregularity. (The OPs argument doesn't show there was.). But we KNOW there was voter suppression and propaganda and enormous amounts of dark money. Let's work on those three known problems. If voter irregularity is confirmed, we can work on that too.

[MichMary]

54. Question about exit polls

I live in a small town, where my polling place probably has at most, a thousand or so voters for a presidential election. I have never been exit polled, which makes sense. Why would anyone stand outside a polling place which has a small number of voters? Well, multiply my insignificant little district by the thousands of other similar districts (most of which probably lean conservative,) and it may more than offset the exit polls of larger cities, where exit polling is probably much more likely to happen, and which will tend to lean Democrat. Could that account for the discrepancy?

[tomp]

59. as I said, I'm no statistician....

....but exit polls were consistent with the vote counts up until GWB. Then electronic voting increased exponentially and the exit polls began to disagree with the vote count. The basic take on this was to discount the polls, not the count. Doesn't make sense to me. I think voting is hacked.

[L. Coyote]

83. Exit polling is inferential statistics, sampling in a scientific manner.

Not every place is sampled. The sample has to be representational. Factors you cite are not going to determine that certain types of places are left out of the sample. That would make the polling unscientific.

Exit polling happens in every sort of place, but only in a sample of them. Your understanding is simply inadequate to discuss the topic.

[Honeycombe8]

60. Professional polls take all that into account.

People lie. But I think it's virtually impossible for millions to lie en masse. People are busy w/their daily lives. They really don't pick up on things like that.

Take some of the interviews I've seen of Trump voters in rural areas. They barely keep up with the news, some not knowing generally a lot about Trump. They would not have known to lie...they live in a little world where they don't keep up with civic events, facebook, etc.

As I recall there were 13 or 14 respected polls. All but one ended up being wrong. Plus, some showed a wide lead by HRC. It would be virtually impossible for all of them to be off that much.

The only one that was correct was a Republican-leaning poll.

That has never happened in our history, since polling began.

Add to that some suspicious things that Trump said before that were red flags, at least to me.

The meme he started about "the election is rigged!" That is straight out of Rove's (and Russia's) playbook...you say that ahead of time, so when and if the opponent says it later, it seems like a copycat thing or defensive, and doesn't have the impact it should.

He was in PA, where HRC had a strong lead. He says the oddest thing: "There is NO WAY I'm going to lose PA! No way! If I do...the election is rigged!" It was so odd that it caught my attention. It made no sense.

Things like that. These things aren't conclusive. But they are suspicious. And nothing can be done about it, now.

The FBI did not look at to what extent the Russian interference was successful. If that's even possible.

[american_ideals]

104. I don't disagree

But I'd add that Putin/Trump didn't really think they were going to win - Putin just wanted to damage Hillary so when she got to office she'd be unable to accomplish anything.
It's hard for anyone to know that 170,000 votes out of tens of millions would fall such that DJT would win.
That said, Putin/Trump did focus on Michigan, PA, and WI.

[Honeycombe8]

106. I think Trump didn't expect it.

At least earlier on.

But I do think it's probably not that hard to zero in on a few select districts in several states, as far as doing certain kinds of nefarious deeds.

We may never know the details. And if Russia had been able to get into some of the voting booths or districts, that changes everything. Any district with machines & no way to print out the votes would be a prime target.

I'm not saying that happened. Who knows?

But near the end Trump did seem to know he was going to PA, and there was reason he should have. HRC was 6 points ahead, and she had the momentum. Yet with his big mouth, he couldn't help but state in his rallies there, "No way I'm gonna PA! No possible way! If I do, the election is rigged! Because I know I'm going to win here!" Just bloviating maybe, but something about it caught my attention at the time. It didn't sound right. It sounded like he KNEW something.

They said they had their own polling, which may explain it. But we now know how incompetent the Trump team is, so it's not likely their polling was more accurate than the 13 professional polls.

[TheFrenchRazor]

73. that's what they said in 2000; they haven't fixed this yet? these polls used to be very accurate...

[Hoyt]

11. +1. And it doesn't do us any good.

[The Wielding Truth]

26. True, but how can any election be fair if their can be no recount and voting

machines have been shown to be vulnerable to hacking?

[SticksnStones]

32. There's been a great deal of forensic review of the data

Reasonable review by educated people. Check out unhackthevote.com

[LonePirate]

56. She used a straight forward calculation of win probabilities from 538 for FL, MI, NC, PA and WI

He threaded the needle to pull off a highly unlikely sweep of those five states. Whether he won legitimately or due to hacking, I cannot say. His chances were low but they were not zero and some people interpret (very) low chances as effectively being zero, which they are not.

[Quixote1818]

57. I agree, I could tell Nate Silver was really unsure about what might happen

All the other models were showing a blow out but Nate saw something concerning and had it a lot closer. He should have stuck stronger with his hunch.

[triron]

6. If you google "Ron Baiman" you will find

a very good analysis of exit poll data that shows a similar improbable result.

[mythology]

17. No you won't find good analysis

Bainman uses "raw" or unadjusted exit poll data which means his conclusions are unmitigated bullshit. Nobody should give that idiot the time of day.

https://www.thenation.com/article/reminder-exit-poll-conspiracy-theories-are-totally-baseless/

https://www.theguardian.com/commentisfree/2012/jun/06/outraged-wisconsin-exit-polls-so-wrong

[LaydeeBug]

40. I will check him out because this seems to be bothering some and they gaslight reflexively

Like...they hacked our rolls...they micro targeted people, they threw people off the voting rolls and gerrymandered. All of this provable fact...but *somehow*, they just *STOPPED SHORT* of hacking or voting machines. *THAT* is just too much, right?

No one with any nefarious purpose would have even TRIED that...we should stop saying it because exit polls are wrong...at least that's what they want you to believe.

Don't let anyone gaslight you.

Each step of the way, they've called us crazy. But here we are...on this path...we've been proven right time and time again.

[kairos12]

8. Those States were Pennsylvania, Wisconsin, Michigan, Florida, what was the other?

[progree]

27. North Carolina. Please see #25 for the percentages and methodology (nt)

[Lucky Luciano]

12. I'm a math guy and that logic is unsound.

They were most likely calculating those probabilities based on each of the five states going to the wanna fuhrer as independent events. They are highly correlated events, so the probability was much much higher and totally feasible - but yes - below 50% by a decent amount.

[sandensea]

16. You better believe it.

And with the easily-hacked read-and-write (Rethuglican-designed) software in most of those voting machines, it must have been a cinch.

[LostinRed]

18. People win the lottery

Math teacher needs to learn probability and I don't know where he gets that math. But remember we won the popular vote!

[eggplant]

20. This argument would make sense if we held millions of lotteries.

And yes, I've learned about probability.

[LaydeeBug]

31. thank you. nt

[eggplant]

63. The funny thing is that I meant to say millions of elections.

But it works either way.

[L. Coyote]

87. When same guy wins lottery five times, probability is very low. That's what the OP is saying.

The OP isn't saying the same guy can't win the Lotto five times. It is possible and you can define the odds by multiplying the odds of winning each one.

[JoeOtterbein]

19. I live in York PA and I have said many times that when I vote I touch a name on the screen and then

who knows what happens? No receipt, no paper, not even a punch on a roll of paper. When I first moved here over 20 years ago we at least had that.

Also, consider that two years ago, in an off-year election, PA elected my fellow proud York, PA citizen, Tom Wolf, for Gov. Tom is a guy as opposite a politician from Trump as anyone can be.

Then we elect Trump.

Not possible for me, unless Russia is involved.

[TheFrenchRazor]

75. the machines are a joke; no one would accept them for anything else important. nt

[keithbvadu2]

21. I'm not so hot on math tonight but wouldn't .01 be one percent and the 16 would be 16 one hundredths

I'm not so hot on math tonight but wouldn't .01 be one percent and the 16 would be 16 one hundredths of that 1 per cent?

[progree]

25. Yes, that math teacher doesn't know percentages

and doesn't know jack shit about probability and statistics either.

0.0016 is 0.16% or 16 hundredths of a percent as you say.

Apparently an elementary school math teacher.

And as others have said, (s)he assumed the events were independent, and of course they are not.

And why did (s)he multiply 5 states together? Why not more? Or less? Why not just the 3 closest ones that if Hillary had won -- she would have been president? Or one could multiply the probabilities of all 50 states coming out as they did -- the result would be a much smaller probability. (And again such calculations would be assuming independent outcomes)

And of course, any calculation, even when done by an inspired mathematical / statistical genius, is only as good as the data that is used. The 538 state probabilities aren't established scientific facts. As they say, GIGO (Garbage In, Garbage Out)

-------------- From Huxenstein's tweets ------------------------------------

I'm a Math teacher. Prob Trump would win all 5 swing states was .0016, 16 ten thousandths of 1%.

Take the probabilities from 538 and turn them into decimals, then multiply

Fl(44.9%=.449)
MI(21.1%=.211)
NC(44.5%=.445)
PA (23.0% =.230)
WI (16.5%=.165)
Multiply together to get .00159, which I rounded to .0016

---------------------------------------------------------------------

[L. Coyote]

88. You obviously know very little about statistics, too little to criticize math teachers.

Statistics expressed probabilities as P = 1.0 if the probability is 1 : 0. P = 0.0016 means the probability is one in a ratio with the inverse of 0.0016. Sorry, this is probably over your head already. P = 0.0016 is equal to odds of one in 625.

[progree]

90. Sorry, a probability of 0.0016 is not 16 ten thousands of 1%. And yes, I have an MSEE

with 4 statistics courses under my belt, including graduate level. And plenty of experience with peer reviewed statistical analysis in my engineering work.

And plenty of others agree with me on this in the thread.

From the tweet: "Prob Trump would win all 5 swing states was .0016, 16 ten thousandths of 1%. "

If you really think that a 0.0016 probability is 16 ten thousandths of 1%, I'm really sorry for you. It is 16 ten thousandths, not 16 ten thousandths of 1%.

To take a simple example, lets say the probability of an event is 0.01 which also is 1/100. That is clearly a 1% probability of that event happening. Would you argue otherwise? Hell, that's middle school math that you convert a proportion to a percent by moving the decimal 2 places to the right.

If you are in awe of elementary math school teachers, than good for you.

>> P = 0.0016 is equal to odds of one in 625 <<

Yes, I agree. Did I ever say or indicate otherwise? A little bit of a strawman argument on your part, perhaps?

It is also 0.16% = 16 hundredths of a percent.

[Liberty Belle]

22. The Republicans could also have hacked the voting machines.

I've thought from day one this is what happened.

[Liberty Belle]

23. Putin probably didn't need to; the Republicans already rigged them in red states

with paperless voting machines as they've been doing for a long time.

[LaydeeBug]

35. Oh, I think the Russians and Republicans coordinated with one another for sure.

I just think it is foolish to examine the expanse of what they did, and then turn a blind eye by pretending the malfeasance stopped with voting machines.

That kind of assholery actually helps Russia.

And I am not about helping Russia at all.

[triron]

66. Same thinking here. Also screws democracy and social progressivism.

[keithbvadu2]

36. from accounting class... If it cannot be audited, something is suspicious.

from accounting class... If it cannot be audited, something is suspicious.

That bullcrap that printers would be too expensive is a total lie.

Printers are not too expensive to put on credit card machines.

[boston bean]

24. Florida seemed most improbable to me.

[LaydeeBug]

34. Check out Mike Farb on North Carolina.

Link to tweet

[L. Coyote]

89. And Wisconsin in relation to black box voting percentage.

Where there are more touch screens, the Trump/Republican red shift goes up.

[58Sunliner]

29. You betcha.

[KingCharlemagne]

37. You should delete this thread. It makes DU look stupid, i.e., suffering

from innumeracy. The plain fact is that precincts and counties that were +12 to +15 for Obama in 2012 swung to +1 to +2 Trump in 2016. Even if Russia did have access to tabulations of ballots, there is no way they could tilt an election by 15-17 points.

Where's the evidence??????

[american_ideals]

42. Voter suppression and propaganda shifted them

Voter suppression, propanganda, and money in politics are what we need to focus on fixing.

[KingCharlemagne]

43. I think Hillary's campaign also suffered from over-confidence. She repeated Gore's

error from 2000 of taking the base for granted. Field workers in 3 of the 5 states mentioned were screaming for more resources on the ground, little or none of which was forthcoming from Hillary's campaign.

Moral: Never, ever take your base for granted.

[SCantiGOP]

45. Thanks for the voice of reason

Time to trash this thread and move on. I agree posts like this one make us all seem silly.

[Awsi Dooger]

65. Hillary repeated Gore's mistakes and added a devastating one of her own

Pursuing North Carolina was lunacy. That state is only going to tumble our way when all the meaningful ones are already secure. I would argue that Obama narrowly carrying North Carolina in 2008 was one of the worst things to happen in Democratic politics in my lifetime, and certainly in regard to the 2016 campaign. It provided a false sense of optimism in that state and totally obscured the meaningful numbers: North Carolina has 43% self-identified conservatives, as revealed in the 2016 exit poll. That means we have virtually no margin for error. The state is 43% conservatives, 22% liberals and 35% moderates. The moderates do tend to vote Democratic more dependably than in other Southern states, so a moderate can be elected like in the governor's race, but that 43-22 split is such disadvantage when compared to the national numbers of 35-26. It simply isn't a true swing state.

These are not independent events, as others have already emphasized. Only a bonehead math specialist wouldn't understand as much. The thread indeed makes this site look very foolish. One of the reasons Nate Silver had Hillary's advantage so comparatively low was that his model identified the potential for widespread small shifts in all the vital states, and if that happened then Trump could pull out an unlikely escape. Check Nate's summaries leading toward election day, especially when he mocked all the 90-99% assertions. He wrote point blank that the polls could be off by a few points, and if so then it likely would be in the same direction everywhere. I had to learn and research that type of thing when I started betting politics heavily and seriously in 1996. It's hardly like sporting events where a theoretical 4 to 1 edge in a game at Denver is not related at all to a theoretical 3 to 1 edge in a simultaneous game in Dallas.

I thought Trump had a 1 in 4 chance on election day. Naturally you don't expect a 1 in 4 to happen. I was relatively relaxed and confident that afternoon, doing some gardening while chatting with my neighbor, a fellow Hillary supporter. Then within hours I was stunned and that neighbor was banging on my door, asking what the hell was going on. I told him it was already over.

2016 was an extremely rare combo. Right wingers are paranoid and distrusting of media and the process to begin with. So it makes sense they won't respond to pollsters in similar dependability as moderates and liberals. We're talking a few percent, not a chunk. Just enough to jeopardize the polling. Since the candidate was such a lowlife with that bus video out there, some of Trump's supporters were embarrassed to advertise that they were still going to vote for him. That applied to family members and trusted friends, not merely to pollsters. Heck, I had it happen within my family and also my circle of friends. Several started chirping that they had voted for Trump, after he won, while either denying the preference or being all but silent in the weeks/months leading to November 8.

Make no mistake, this was an unforced error. Hillary didn't sense the degree of unrest and fear among working class whites so her messaging wasn't loud or effective enough, and consequently she took the midwestern states for granted as if they weren't swing states as all. One of the few non-lies that Trump has told in years is that Hillary campaigned in the wrong states.

[KingCharlemagne]

68. Case in point: her campaign ran TV ads in the SoCal market in

The two weeks before the GE. The genius who came up with that brilliant tactic should never work in Democratic Party politics again.

[Dem2]

76. I know you're right

Because NH. It NEVER should have been as close as it was. These were the same voters that barely put the rodent over the line in the other swing states.

[TheFrenchRazor]

77. YOU should delete your post. where is the evidence that the vote counts were legit? oh yeah, the vot

the votes are invisible, so it's anybody's guess if the results were accurate. i'm sorry, but i don't trust the GOP/ruskies nearly as much as you do. we need all paper ballots, all hand/eye counted, all the time. wake up, it's 2017; computers can be and are hacked every day.

[KingCharlemagne]

85. Written by someone with ZERO understanding of statistics. - nt

[L. Coyote]

94. You should delete this post. It makes DU look like people don't read the OPs.

[Demsrule86]

46. I have always believed Trump is illegitimate and got there by cheating.

He took just enough...and votes were 'lost' in Detroit...PA, MI and WI all had numbers run up in rural areas.

[treestar]

47. I've always wondered about that

Seems like Donald got so lucky.

It's still hard to believe he got PA.

All the right voters in just the right places. Incredible luck.

[dlk]

48. Of Course the Voting Machines Were Hacked

It doesn't take a rocket scientist...

[Enoki33]

61. Not a statistician. The question of hacking of certain types of voting machines

is a real possibility. The orchestrated voter suppression is reality. The bottom line is "we wuz robbed" to quote a baseball player.

[roomtomove]

49. ALL of our elections are rigged...

by voting machine fraud/tampering and manipulating the tabulation. The media and the politicians are afraid of the consequences if this is revealed/made public.

[Honeycombe8]

58. Shhhh. We're not supposed to say that.

It's obvious, of course.

But we're not supposed to say this, since that would result in a lack of faith in the election system, I've read.

But we can't have a discussion about it, if it can't be discussed. I thought this from election night, based on various things, including some things that Trump said BEFORE the election.

[Maraya1969]

62. I've been saying this since the election results came in but couldn't figure

out the exact numbers. I knew the probability of all the stat models, early voting polls, exit polls to be wrong was teeny tiny.

Remember how we were all celebrating the day before the election before the results came in? And then boom!

That is election hacking 101

[Towlie]

64. The number .0016 is NOT "16 ten thousandths of 1%."

Actually, it's 16 hundredths of one percent.

That would still be a compelling figure, but the gaff casts doubt upon the competence of this alleged math teacher and the reliability of the probability calculation. We have plenty of solid evidence against Trump, but spreading this around will only weaken our case.

[SCantiGOP]

69. Absolutely

Really can't understand why your statement doesn't cause this thread to disappear.
Junk science is as objectionable as fake news.

[L. Coyote]

95. True. P 0.0016. Where P 0.50 is 50% probability, i.e. a coin toss.

[drm604]

67. I think absentee votes need to be looked at.

Why were Russian hackers interested in voter registration files?

[ehrnst]

71. Those of us that supported her weren't fooled. (nt)

[paleotn]

72. Oh, good lord.

and one wonders why our nation is so weak in math and science.

[ecstatic]

74. But nobody can or will do anything about it.

What's to stop it from happening again in 2018?

[BzaDem]

96. This is absurd, and shows a basic lack of understanding of high school probability and statistics.

For those who think the results were tampered with, were you making the same argument when Obama overperformed his poll results in 2012 by more than Trump overperformed his in 2016?

This math teacher (really?) is assuming that the polls of the various swing state accurately sampled the population, and were only subject to sampling error (margin of error). In such an imaginary universe, the probability of the errors all going in one direction is very low.

But in the actual universe we live in, there is no easy way to get an accurate random sample of voters. 90% of people polled do not answer their phones. (Think about it -- when was the last time you picked up the phone to an unknown number?) Many people who do answer the phones say they are likely to vote, but don't end up voting. Pollsters make all sorts of assumptions when deciding how to weight the resulting data, and these assumptions are often very wrong. Yet making assumptions is the best they can do, because 90% of the population doesn't respond, and people do not accurately predict their likelihood to vote.

The resulting modeling error (different than sampling error) is that the polls are often off by several points. It is impossible to know before the election what direction the polls will be off, or by how much. But unlike with sampling error (where being off in either direction is equally likely, allowing the error to be reduced by averaging many polls), it is very likely that modeling error will cause most polls to be off in the SAME direction. After all, if (say) non-college educated white voters were marginally less likely to respond to pollsters in Ohio, why in the world would this not also be the case in Wisconsin (or other surrounding states)?

This is precisely why people on election night knew Clinton was in profound trouble once it became clear how she was doing in Florida. Florida was not remotely a must-win state for Clinton. But if she was underperforming her polls in Florida by several points, it was very likely she would do so in the midwest as well. Sure enough, that's exactly what happened.

In this case, the modeling error was a slight undersampling of non-college-educated white voters. This had a decisive outcome on the election, because the swing states in question were primarily in the midwest (with a higher-than-average percentage of non-college-educated white voters). But there were also many non-swing states where Trump over-performed his polls, for the same reason. The idea that Putin or anyone else hacked large numbers of votes in totally irrelevant states is ludicrous.

[L. Coyote]

100. Wrong polls. This is about the exit polls.

Of course, the exit polls and the pre-election polls in the battleground states are in agreement. And the red shift in the exit polls exceeds the pre-election polling and is a multiple of previous election's red shift in the exit polls, which was considered impossible and alarming too.

In the rest of the civilized world, statistics still works perfectly fine. Only in the United States does stati=stics either not function the same as elsewhere, or we have an election rigging problem. Some people seem to prefer to believe that statistics does not work inside the United States and contrive as many excuses to explain it away rather than admit that the United States' voters are misrepresented by the current Republican Party.

Which are you? Do you think statistics is flawed in the USA, or do you think elections were rigged?

[progree]

101. Those percentages in Hux's tweets are not based on exit polls

/-------------- From Huxenstein's tweets ------------------------------------

I'm a Math teacher. Prob Trump would win all 5 swing states was .0016, 16 ten thousandths of 1%.

Take the probabilities from 538 and turn them into decimals, then multiply

Fl(44.9%=.449)
MI(21.1%=.211)
NC(44.5%=.445)
PA (23.0% =.230)
WI (16.5%=.165)
Multiply together to get .00159, which I rounded to .0016

--------------End From Huxenstein tweets ----------------------------------

I looked at the Florida, Michigan, and North Carolina pages at fivethirtyeight.com, and the probabilities match up to Huxenstein's tweet

https://projects.fivethirtyeight.com/2016-election-forecast/florida/
https://projects.fivethirtyeight.com/2016-election-forecast/michigan/
https://projects.fivethirtyeight.com/2016-election-forecast/north-carolina/

Later: ditto the Pennsylvania and Wisconsin pages.

I didn't see any exit polls included.

I don't think the 2016 exit polls factored into it at all, given Nate's attitude about them:

Ten Reasons Why You Should Ignore Exit Polls, 538, 11/4/08
https://fivethirtyeight.com/features/ten-reasons-why-you-should-ignore-exit/

--------------------------------------------------------------------------------------------
As for correlation, here is what Nate has to say:
https://fivethirtyeight.com/features/a-users-guide-to-fivethirtyeights-2016-general-election-forecast/

/----------- Begin excerpts-----------------------------------------------
The error from state to state is correlated. If Trump significantly beats his polls in Ohio, he’ll probably do so in Pennsylvania also. Figuring out how to account for these correlations is tricky, but you shouldn’t put too much stock in models that don’t attempt to do so. They’ll underestimate the chances for the trailing candidate if they assume that states are independent from one another.

... here’s a correlation matrix drawn from recent simulations. You can see the high correlation between Wisconsin and Minnesota, for example.
------------- End excerpts --------------------------------------

There are a lot of high correlations. So the idea that the state probabilities are independent is, well, way out there. I'll take Nate Silver over any elementary math teacher who doesn't know how to convert a number to a percent.

And given that Nate has those probabilities for the 5 swing states shown in the Huxenstein tweet above (and similarly for most of the other swing states), how comes Nate had a 28.6% probability of Trump winning the presidency ( https://projects.fivethirtyeight.com/2016-election-forecast/?ex_cid=2016-forecast-analysis ). A couple of orders of magnitude higher probability than Huxenstein has?

As for polling being more accurate elsewhere than here, I haven't seen that one. How did the Brexit vote turn out?

I don't discount red-shift shenanigans. I just think the error-ridden silly OP (including Huxenstein's following tweets) is a horrible way to advance that argument.

[BzaDem]

102. You are confusing "exit polls" with "statistics". Exit polls use statistics, but a problem with

exit polls does not necessarily imply a problem with statistics (and does not necessarily imply fraud).

Here's an analogy. Computers require electricity to run. But if there is a problem with my computer, that doesn't necessarily mean there is a problem with my electricity, or a problem with the physical laws governing electricity. It also does not imply sabatoge. The electricity could be fine, and the problem could be elsewhere. In fact, one would look pretty silly if they immediately jumped to the conclusion that the most likely explanation is a change of the laws of physics surrounding the computer, or immediately jumped to the conclusion that the problem is sabatoge.

Likewise, if your immediate conclusion when exit polls are off is to assume either statistics no longer apply, or that there was fraud, a re-evaluation of the thought process that led to such a conclusion may be in order.

As a sanity check, exit polls had Gore winning Alabama and Texas. Do you think those were stolen?

Assuming you acknowledge that Gore didn't win Alabama or Texas (or many other states the exit polls had him winning, that he lost by a comfortable margin and that no one expected him to win), it makes sense to dig deeper to examine why the exit polls might be off.

One possible reason might be that exit polls in this country are not conducted to determine the winner. That is not what the exit pollsters claim their data will do, and it is not what they design their surveys to do. A survey that aimed to verify the winner would involve much more funding, many more interviewers, and many more polling places selected for interviews. And even then, certain types of differential non-response bias cannot be corrected for unless known in advance (which is often not the case).

As for why exit polls are supposedly spot on in other countries, this is partly a myth that continues to be perpetuated without critical examination. For example, the 2015 UK general election resulted in Tories gaining a majority, which the exit polls did not predict. In fact, they even have a name for this effect, going back many elections: the shy-Tory voter. For another example, the media's projections on election night in France this year were more or less correct, and people continue to cite this as an example of exit polls being accurate. But looking just a bit behind the curtain, one would discover that these were not just exit polls, but included a full hour of real results from polling places that closed earlier than other polling places still open. The reason the projections were not released from exit polls data alone is that no one reasonably expects the raw unadjusted exit poll data to be accurate with high confidence.

This is not to say exit polls aren't *better* in some other countries in some elections. Exit polls in other countries are often better funded, conducted with many more interviewers in many more precincts speaking with many more (and more representative) voters. If you would like more details on the differences between exit polls in the US and other countries, this is a pretty good resource:

http://www.mysterypollster.com/main/2004/12/what_about_thos.html

[Midwestern Democrat]

99. Explain Minnesota - a state Hillary narrowly WON (by 1.5%). When Minnesota's that tight,

the results in the rest of the Upper Midwest don't look that shocking.