Originally published at http://www.commondreams.org/views/2019/10/30/nate-silver-making-he-goes
All that sparkles is not gold, nor "Silver, " either.
Nat Silver is worse than a weatherman who only looks out his window to predict the weather.
Why he is asked predict/poll anything other than this seasons horrible fashion sense is beyond me.
The arrogance of “analytics”
As someone who actually understand statistics, this article is nonsense. Nate Silver has made some bad predictions (many of which he owns up to), but 2016 wasnt one of them. He gave Trump basically a 1 in 3 shot to win. The author and other like him interpret to mean since Trump won, Nate was 2/3 wrong or wholly wrong. That isn’t how it works. Nate actually has pretty extensive articles at 538 talking about how they gauge if their estimates/predictions are good…and the basic idea is when they say someone has a 1 in 3 chance of winning, they shocker win about 1 in 3 times. Dear author, go get a minor in stats (or a major even) and you’ll say a lot more accurate things.
A poll is only as good as the moment it is taken.
How did Ol’ Nate do in the Congressional races of 2018? The Blue Wave election of Democrats in the House and Ol’ Nate missed by 13. He’s not a weather vane by any means. And, that wave wasn’t a sneaker and didn’t take a stat degree, imo.
Looking back all I missed by was 2, including that special election redo in N. Carolina. Which had problems the 2nd time, as well. And, like Will Rogers said, " all I know is what I read in the newspapers, on the internet, the TV and here at CD.
When you listen to the street corner talking it sometimes, statistically speaking, pays to just listen and take note.
From Jacob Bacharach’s article:
Now, I do not think that Silver is a secret Trumpist; he is just another commentator who, sheltered by his money and network of contacts, is free to view political alignments as mere fandom and political contests as another competitive league full of “stories,” of rivalries and upsets, of failed Hail Mary passes and successful buzzer-beaters.
Which is exactly how the US MSM covers the news: pageantry and personality. Once the conventions are over next year, it’ll be a “two-horse race to the finish line.” In this context, Nate Silver’s transition into the political punditocracy is hardly surprising.
Tim Ryan wrote: “As someone who actually understand statistics… Nate Silver has made some bad predictions…”
He has indeed, and some ‘good ones’ that are likely the result of random chance, and not his ‘secret formula.’
Nate’s central problem- common among would be polling aggregators, is that he violates a very basic tenet of modelling: GIGO. If you put garbage data from cheap media polls and other skewed or methodologically suspect source in- you’ll get garbage out, no matter how diligently you attempt to correct for ‘house effects’ or apply weightings to poorly drawn samples.
I think that the real problem with Silver and any other poll analysts is that polling responses are massively down compared to the 60s and 70s–only 6% on average compared to 70-80% then. That means accuracy is vastly more expoensive than it used to be.
Nate Silver and 538 are dead to me for having the gall to poll superdelegates even this early and report those results. I haven’t looked at the site in many months after seeing that. His comments about Sanders are just more evidence he has zero value to progressives.
eridani wrote: think that the real problem with Silver and any other poll analysts is that polling responses are massively down compared to the 60s and 70s–only 6% on average compared to 70-80% then. That means accuracy is vastly more expensive than it used to be."
There is that, too -although we should note that accurate research- i.e., getting a representative sample, has always been expensive, which is why, all else being equal, cheap media polls have always been suspect.