|
EugeneJ posted:Nate Silver vs. Karl Rove in a game of chess Silver wins; he's actually a smart guy, just someone who was seriously wrong on Trump, and has been backpedaling clumsily since then instead of just owning it.
|
# ? Feb 26, 2016 06:40 |
|
|
# ? May 4, 2024 02:46 |
|
Just came across this in The Atlantic:quote:Polls in individual states tell a similar story. In a recent piece on FiveThirtyEight, Nate Silver noted that “about half” of Republican primary voters in New Hampshire said that “they’d be unhappy with [Trump] as their nominee.” That’s true: 46 percent of New Hampshire Republican voters told exit pollsters they would be “dissatisfied” if Trump won the nomination. What Silver didn’t mention is that the percentage that said they’d be dissatisfied if Rubio or Cruz won was even higher. http://www.theatlantic.com/politics/archive/2016/02/the-ceiling-buster/470919/
|
# ? Feb 26, 2016 06:52 |
|
Sebadoh Gigante posted:Just came across this in The Atlantic: Ugh where the gently caress do I go now to get my 538 fix now
|
# ? Feb 26, 2016 08:13 |
|
Re-posting my criticism of Nate Silver from another discussion: "As someone who uses statistics on a daily basis in scientific research, reading Silver's forecasting book was a bit sad. He has a good grasp on a limited subset of statistics, but then tries to apply it everywhere, without considering how appropriate it is. Towards the end he totally goes off the deep end with claims of being able to predict terrorist attacks. In short, Nate is not the forecasting god he makes himself out to be, and falls into the trap of weighing his own past successes too heavily, something he warns against in his own book, but doesn't apply to himself." Silver bought into his own hype and stopped following his own principles of prediction a while ago with this race.
|
# ? Feb 26, 2016 10:09 |
|
unskew my neg rear end.
|
# ? Feb 26, 2016 13:28 |
|
Is this where all the berniebros feverishly unskew Nate Silver in a totally-not-pathetic way to explain why their 1000 year old candidate isn't poo poo?
|
# ? Feb 26, 2016 22:54 |
|
Move over Nate Silver there's a new kid on the block, Trump Will Become President, Says Extremely Accurate Statistician
|
# ? Feb 26, 2016 23:37 |
|
Goatman Sacks posted:Is this where all the berniebros feverishly unskew Nate Silver in a totally-not-pathetic way to explain why their 1000 year old candidate isn't poo poo? We're talking about trump here you ..... Maroon
|
# ? Feb 26, 2016 23:44 |
|
Squalid posted:It would not be an outlier. As has been mentioned, the sample is really too small to make that kind of a conclusion. As any loyal 538 reader should remember, the theory that the party decides has already stumbled a few times, in particular endorsements in the 2008 Democratic primary pointed to Hillary up until the end of the race. A Trump win would seriously challenge the underlying assumptions. Yeah, Trump's not so much an outlier as an out-of-sample event. I won't claim that he's singular, but he doesn't play politics by the same rules as everyone else, which means that models built on the notion that standard "political causation" will hold shouldn't be used. In those cases, I'd want a model that makes the fewest assumptions about the underlying politics, which may just be an averaged set of polls rated by quality. After all, that popularity leads to election wins is a relatively uncontroversial assumption. It's not surprising that Nate's 'polls-plus' model seems to be completely imploding. He seems to be coming around to that view too: https://twitter.com/NateSilver538/status/703321897283493888 CottonWolf has issued a correction as of 23:55 on Feb 26, 2016 |
# ? Feb 26, 2016 23:50 |
|
Mr.48 posted:Re-posting my criticism of Nate Silver from another discussion: Seconding this. i just started something very much inspired by 538 (shameless plug: http://www.nanaya.co) - between that and science/academia it's easy to fall into the trap of drinking your own cool aid as you realize early modeling works. it only takes a few new cases to unmistakably break a theory. Unfortunately for presidential elections, it's not like there's that many samples to learn from. Of course you can reapply the theory to historic data which I've never seen them do. Also, is it just me or has their modeling methodology become a lot more opaque or just dumbed down to the point where it is: http://fivethirtyeight.com/features/how-we-are-forecasting-the-2016-presidential-primary-election/ In a multifactor model independent weighting makes a huge difference - they don't really describe that...
|
# ? Feb 27, 2016 00:40 |
|
Dear Nate please fire the failed comedian who tries too hard in the significant digits articles and bring back Mona thx
|
# ? Feb 27, 2016 02:55 |
|
Did everyone see my ingots joke?
|
# ? Feb 27, 2016 03:10 |
|
The Revenant posted:Did everyone see my ingots joke? It was kinda clever, OP
|
# ? Feb 27, 2016 03:12 |
|
Dear Nate I've hated your new site ever since the head shots with the weird colors came out. Makes you look super pretentious. K byyyye
|
# ? Feb 27, 2016 05:21 |
|
I think the fundamental problem has been an under-recognition of the importance of mechanisms - the "how" instead of the "what". And so when those mechanisms broke down or were circumvented by the Donald, that was somewhat ignored. Folks just stuck with saying "The Party Decides" to assume that Trump would get stumped, without really using that as a framework to look into how the GOP might go about the stumping or whether they remained on track, etc.G-Hawk posted:http://fivethirtyeight.com/features/donald-trumps-six-stages-of-doom/ See, I completely disagree - this article is great not because it's accurate, but because it fully lays out the thinking. I can read the article, see exactly where Silver's coming from, and then adjust that logic to my own beliefs to come to my own conclusions.
|
# ? Feb 27, 2016 20:06 |
|
Zoran posted:It was kinda clever, OP tahnk u
|
# ? Feb 27, 2016 20:18 |
|
i hope "the party decides" really does die after this election because its a pretty lovely and reductionist way to look at primaries
|
# ? Feb 27, 2016 21:00 |
|
Jackson Taus posted:See, I completely disagree - this article is great not because it's accurate, but because it fully lays out the thinking. I can read the article, see exactly where Silver's coming from, and then adjust that logic to my own beliefs to come to my own conclusions. maybe it would be more palatable to most people if it was just a clear op-ed instead of assuming every event has a 50% chance of occurring and pretending to have an article based in Stats and Real Metrics and Objective Analysis that being said, no idea how he could smokyprogg has issued a correction as of 21:33 on Feb 27, 2016 |
# ? Feb 27, 2016 21:17 |
|
babypolis posted:i hope "the party decides" really does die after this election because its a pretty lovely and reductionist way to look at primaries Agreed.
|
# ? Feb 27, 2016 21:50 |
|
Continuing to use polls-plus months after dozens of endorsements for Jeb! failed to move his polling numbers at all was a pretty bad look.
|
# ? Feb 27, 2016 22:06 |
|
Silver was very good when he was cutting through the crap and going straight to the data. Problem is, that gets boring and anyone else can do it too, so he's tried to do something better. The issue is, that something better is the same sort of entrail reading everyone else was doing when he did his idea of lets just look at the actual polling data thing. 538's model is garbage this year and he should abandon it rather than trying to justify its failures. Plus the analysis the site produces these days is clickbait garbage.
|
# ? Feb 27, 2016 22:10 |
|
the jelly nun posted:It's hilarious to me that people can't see Nate Silver and his website have become paid shills for the highest bidder. Wow, what a coincidence that this guy stayed away from punditry and predicted everything correctly 2 elections in a row, then right after his website becomes some ESPN Disney thing he suddenly turns into a dick Morris level moron barely less embarrassing than Jeb Bush, who posts article after article sucking GOPe and Clinton dick. Toilet Mouth posted:This post is best read in Dennis Hopper's voice.
|
# ? Feb 27, 2016 23:04 |
Pop quiz, Hillary, should women alleging that they were raped by a powerful man be shamed into silence?
|
|
# ? Feb 28, 2016 00:09 |
|
KaptainKrunk posted:Continuing to use polls-plus months after dozens of endorsements for Jeb! failed to move his polling numbers at all was a pretty bad look. or how rubio had every single important endorsement in south carolina and still got destroyed by trump
|
# ? Feb 28, 2016 00:53 |
|
babypolis posted:i hope "the party decides" really does die after this election because its a
|
# ? Feb 29, 2016 01:50 |
|
Concerned Citizen posted:The journey of Nate Silver, Data Journalist and Trump Skeptic, in one twitter feed:
|
# ? Feb 29, 2016 02:16 |
|
reignofevil posted:Look at this guy freaking out over an outlier. it's not my theory having the problems duder
|
# ? Feb 29, 2016 02:24 |
|
my favourite political science factoid is that the rate of false positives you'd expect in an entirely random world corresponds roughly to the acceptance rate of academic papers in poli sci journals
|
# ? Feb 29, 2016 02:25 |
|
Kazak_Hstan posted:His dad and brother? If the fact that they were related to them reduces the impact of the effect, that's actually really really strong evidence Nate has the causation chain backwards.
|
# ? Feb 29, 2016 03:01 |
|
GlyphGryph posted:If the fact that they were related to them reduces the impact of the effect, that's actually really really strong evidence Nate has the causation chain backwards. IIRC, 538 only factors in the endorsements of currently serving elected officials. also at the new thread title. Nate Silver is a very low-energy statistician. Nate is a mess. Nate is a waste. Nate is a big, fat mistake. He's a lightweight, and an even lighter-weight when you apply the polls-plus model.
|
# ? Feb 29, 2016 03:14 |
|
As wrong as he's been about Trump, it looks as if he was mostly right about Sanders.
|
# ? Feb 29, 2016 04:35 |
|
punk rebel ecks posted:Pretty much this. Nate Silver embarrassed himself. At the end of the day immediate numbers can't tell you everything. This election was like none other and it isn't fair to compare it to previous ones. Anyone who even somewhat paid attention to the primaries would be able to tell that Trump was a force to be reckon with.
|
# ? Feb 29, 2016 05:37 |
|
Zohar posted:my favourite political science factoid is that the rate of false positives you'd expect in an entirely random world corresponds roughly to the acceptance rate of academic papers in poli sci journals Lol
|
# ? Feb 29, 2016 05:56 |
|
Frosted Flake posted:As wrong as he's been about Trump, it looks as if he was mostly right about Sanders. 50/50 between two candidates? What are the odd?
|
# ? Feb 29, 2016 05:58 |
|
G-Hawk posted:The irony is that if Nate had just looked at numbers, he would have been less bearish on Trump. Instead he spent 6 months finding any way possible to discount polls, the numbers. This is also true. While the numbers weren't spot on, they did show some holes in his "conclusion". The most notable was how it was pretty clear that Trump was the candidate who held his popularity, even if he wasn't number 1 all the time, rather than being your the flavor of the month candidate like say Carson or Cruz.
|
# ? Feb 29, 2016 08:12 |
|
Frosted Flake posted:As wrong as he's been about Trump, it looks as if he was mostly right about Sanders. Not really. He was right in the sense that, of the two candidates, he correctly predicted that the frontrunner would win the nomination. Which isn't exactly a challenging pick. But he pointlessly dismissed the surge in support that happened in January. There was a brief period in time where it looked like Bernie might have a very solid shot at winning the nomination.
|
# ? Feb 29, 2016 08:24 |
|
Nate Silver and other data journalists were basically running on the theory that candidates would boom-and-bust like Herman Cain and other candidates in the 2012 race. They just sort of applied 2012 as a universal model for no particular reason, and ignored the reasons why 2012 happened the way it did in the first place (because the right disliked Romney). When the bust never happened, I guess they just sort of assumed that the crash would be all the more epic later on.
|
# ? Feb 29, 2016 08:26 |
|
Concerned Citizen posted:Not really. He was right in the sense that, of the two candidates, he correctly predicted that the frontrunner would win the nomination. Which isn't exactly a challenging pick. But he pointlessly dismissed the surge in support that happened in January. There was a brief period in time where it looked like Bernie might have a very solid shot at winning the nomination. His take was that Sanders could win Iowa and New Hampshire and lose everything else, which is basically how it looks now.
|
# ? Feb 29, 2016 08:35 |
|
Concerned Citizen posted:Nate Silver and other data journalists were basically running on the theory that candidates would boom-and-bust like Herman Cain and other candidates in the 2012 race. They just sort of applied 2012 as a universal model for no particular reason, and ignored the reasons why 2012 happened the way it did in the first place (because the right disliked Romney). When the bust never happened, I guess they just sort of assumed that the crash would be all the more epic later on. I noticed that, and commented on how, if anything, Trump's pattern of support was more like a supercharged Romney than one of the not-Romneys.
|
# ? Feb 29, 2016 08:35 |
|
|
# ? May 4, 2024 02:46 |
|
Bip Roberts posted:His take was that Sanders could win Iowa and New Hampshire and lose everything else, which is basically how it looks now. Well, prior to Hillary's polling surge last week Bernie was on track for possible wins in MA, CO, MN, VT, and maybe even OK. So I think that fundamental thesis was wrong - and 538 was so overweighting the wrong factors (like endorsements) that it gave Hillary a 58% chance of winning NH a few weeks out from the primary.
|
# ? Feb 29, 2016 08:41 |