US election 2016: statistician Nate Silver's big Donald Trump mistake

The Sydney | October 31, 2016 | Margaret Sullivan

[2ndDivisionVet]

Link only due to copyright issues: http://www.smh.com.au/world/us-election/us-election-2016-statistician-nate-silvers-big-donald-trump-mistake-20161030-gseaye.html

[ez]

Let’s have the fingerprint of the voter on the ballot too in ink that takes at least 24 hours to fade.

[InterceptPoint]

Nate Silver has a very good methodology with regard to his statistical analysis. But it is garbage in and garbage out when he uses polls that are way over weighted in terms of D/R/I and then corrects them only a point or two.

He should know better. He does know better. And he knows how to make the necessary corrections. But, I have to assume that he actually believes that the Dems will outvote the Republicans by 7 or 8 points. Those are the sort of numbers the pollsters are using and some are worse than that.

But I still expect Little Nate to tell us that the race is "growing tighter" and that the latest email scandal is "giving Trump momentum" as he moves the needle back toward a 50/50 chance for Trump and Hillary next Tuesday. That's what the garbage polls will be doing and he will follow.

[MUDDOG]

Krugman’s a pet peeve of mine too. Big time.

[edh]

“Krugman’s a pet peeve of mine too. Big time.”

He is disgusting. He parades around as an economist when he’s nothing more than a political hack. That alone makes me want to vomit. Hack is too kind of a word, but I will leave it at that :-).

[ModelBreaker]

“Nate silver is going to be to the Democrats in 2016 what Dick Morris was to the Republicans in 2012. Big difference is dick’s predictions were not until the day of the election. Nate’s predictions have been All year.”

One big difference. Nate Silver has been right at least one time. Morris just tells people what he thinks they want to hear so they will buy books.

[Vision Thing]

Hey, Nate writes books, too.

The Signal and The Noise is wildly entertaining. Especially in the chapter where Nate and Donald Rumsfeld reveal their mutual admiration for one another.

But don’t worry, I didn’t buy the book. I borrowed it from the library instead.

[James Thomas]

Great pluck, 2ndDivisionVet.

[Pollster1]

The second, more important one is basing his predictions on technical analysis (in financial terms) rather than fundamentals.

Overly technical commentary:

Nate Silver's analysis is based on the assumption that the polls are unbiased. His probabilities essentially come from a Monte Carlo simulation drawn from a population that would produce the observed poll numbers. If the polls are biased, then Nate Silver's probabilities are useless. In this case, we all know the polls are biased. Even if pollsters wanted accurate results, polling response rates have dropped dramatically, and we don't know the extent of the drops among those who support Trump or among those who do not.

[Red Steel]

Little Natie is slow on the uptake. 8 days out from the election here he is not certain anymore.

- - - - -

“Nate Silver Verified account
@NateSilver538

Either polls are having trouble capturing a random sample, or voter preferences are quite volatile. Either way, lots of uncertainty.”

- - - - -

Silver what? Don’t you have Trump at 98% chance of losing? A maroon.