|
Cingulate posted:Well to be fair, he is a poll aggregator, not a reality aggregator, and it's not clear how early voting actually behaves cause it's current prevalence is a recent development. Right? Absolutely, but I'd argue that early voting information is basically a poll in and of itself- it's just measuring how many people of each party affiliation are Actually Voting. We've had a few cycles now of it, and we can make some decent assumptions about what a typical early voting electorate is like compared to an election day electorate. We can also weight the early voting results by what fraction of the electorate's already voted, too. It's just silly that a model would show Nevada as still a 50-50 shot when we know some absurdly high fraction of people have already voted, their party affiliations are known, and unless there's a number of Democrats For Trump that is wholly unprecedented then Trump would have to run up the score to an extent nobody's ever seen on election day to win. It's not 50-50. Is it 80, 90, 95? I dunno. I am not familiar enough with the data. But that's one area I'm confident 538's model is badly overestimating uncertainty because of not using data.
|
#
?
Nov 5, 2016 22:17
|
|
|
|
| # ? Jun 11, 2024 06:47 |
|
|
Cingulate posted:Well to be fair, he is a poll aggregator, not a reality aggregator, and it's not clear how early voting actually behaves cause it's current prevalence is a recent development. Right? His argument for not including early voting is that he sort of indirectly does. quote:We get a lot of questions about why our model doesn’t account for early-voting data. One answer is that it does, to the extent that early voting is reflected in the polls. Another is that the whole point of building a model is to take a more disciplined approach toward evaluating evidence instead of just throwing a hodgepodge of indicators together. Our model does a great job of reflecting what the polls say and translating that into probabilities. So you can take it at face value or use it as a departure point if there are some other factors you might want to consider. But there haven't really been any reasonable polls in Nevada since then. And there is still the problem that Latinos are underrepresented in polls. cant cook creole bream fucked around with this message at 22:20 on Nov 5, 2016 |
|
#
?
Nov 5, 2016 22:17
|
|
|
Spaced God posted:https://twitter.com/RawStory/status/794986090033717248 Good grief. The consent decree better get extended after this is all over. Also that lawsuit better not get dismissed just because the guy stepped down after that. Man. This is terrible. At least he was stopped relatively quickly.
|
|
#
?
Nov 5, 2016 22:21
|
|
|
RVProfootballer posted:I only glanced at this, but isn't that first link just correlations between results in the simulations? As in, it takes into account polls, correlation between states, and everything else? The actual modeled correlation between states can't be that high, I don't think, else the model would have Iowa and Ohio very likely to go to Clinton given polling in Wisconsin, Minnesota, and Michigan (I suspect, anyway). But he basically just says "it's complicated, but we account for state dependence when estimating errors don't ask how it's complicated". cant cook creole bream posted:No. Polls plus only puts in a few extra points based on the economy and the fact that it would be the third democratic presidency in a row since that's always less likely. By this point in the game polls only and polls plus have basically converged. (as they are supposed to.) Mukaikubo posted:Absolutely, but I'd argue that early voting information is basically a poll in and of itself- it's just measuring how many people of each party affiliation are Actually Voting. We've had a few cycles now of it, and we can make some decent assumptions about what a typical early voting electorate is like compared to an election day electorate. We can also weight the early voting results by what fraction of the electorate's already voted, too. It's just silly that a model would show Nevada as still a 50-50 shot when we know some absurdly high fraction of people have already voted, their party affiliations are known, and unless there's a number of Democrats For Trump that is wholly unprecedented then Trump would have to run up the score to an extent nobody's ever seen on election day to win. It's not 50-50. Is it 80, 90, 95? I dunno. I am not familiar enough with the data. But that's one area I'm confident 538's model is badly overestimating uncertainty because of not using data. This is a very academic perspective though.
|
|
#
?
Nov 5, 2016 22:21
|
|
|
Cingulate posted:Does anyone have a convincing argument for why it would be in Nate's interest to make the prediction look worse for Clinton, rather than better for her? millenialmotivation.txt
|
|
#
?
Nov 5, 2016 22:22
|
|
|
Mukaikubo posted:Absolutely, but I'd argue that early voting information is basically a poll in and of itself- it's just measuring how many people of each party affiliation are Actually Voting. We've had a few cycles now of it, and we can make some decent assumptions about what a typical early voting electorate is like compared to an election day electorate. We can also weight the early voting results by what fraction of the electorate's already voted, too. It's just silly that a model would show Nevada as still a 50-50 shot when we know some absurdly high fraction of people have already voted, their party affiliations are known, and unless there's a number of Democrats For Trump that is wholly unprecedented then Trump would have to run up the score to an extent nobody's ever seen on election day to win. It's not 50-50. Is it 80, 90, 95? I dunno. I am not familiar enough with the data. But that's one area I'm confident 538's model is badly overestimating uncertainty because of not using data. I'm pretty sure that's right. But, is there really reliable demographic split of early voters versus on the day voters at this point? Knowing very little about US political breakdowns, I'm not sure, but purely from a sample size perspective it sounds like it would be horrifying to get a good estimate of, especially if it differs systematically between states.
|
|
#
?
Nov 5, 2016 22:23
|
|
|
Why do you even need to intimidate voters in loving Arkansas?
|
|
#
?
Nov 5, 2016 22:25
|
|
|
It seems very unlikely that Nate Silver is intentionally favoring the Republicans (rather more likely that he is intentionally using a very high uncertainty, but whatever), but it's almost certain that Nate Silver knows how few of his audience realize that his model predicts Clinton winning about half of the light red colored swing states.
|
|
#
?
Nov 5, 2016 22:26
|
|
|
James Garfield posted:It seems very unlikely that Nate Silver is intentionally favoring the Republicans (rather more likely that he is intentionally using a very high uncertainty, but whatever), but it's almost certain that Nate Silver knows how few of his audience realize that his model predicts Clinton winning about half of the light red colored swing states. * deliberately ignoring the punditry. Not gonna defend that.
|
|
#
?
Nov 5, 2016 22:29
|
|
|
nachos posted:Why do you even need to intimidate voters in loving Arkansas? For the same reason Trump's campaign is sending poll observers to SC--they're not very smart. You could run Gay Lucifer on the GOP ticket in SC and the evangelicals would fall over themselves to endorse red-dick sodomy.
|
|
#
?
Nov 5, 2016 22:30
|
|
|
ReV VAdAUL posted:People tend to prefer a nice surprise to a nasty shock so being too pessimistic probably wouldn't hurt him too much. It won't matter that gaining FL, for example, when he says it's for Trump is better than losing FL when he says it's for Clinton. What matters is his aggregate system is saying Trump will win it and the others' say Clinton will and he'll lose prestige and relevance because he earned it in the first place by being the most accurate predictor.
|
|
#
?
Nov 5, 2016 22:31
|
|
|
CottonWolf posted:I'm pretty sure that's right. But, is there really reliable demographic split of early voters versus on the day voters at this point? Knowing very little about US political breakdowns, I'm not sure, but purely from a sample size perspective it sounds like it would be horrifying to get a good estimate of, especially if it differs systematically between states. But if they are overwhelmingly showing a lead, it's unlikely that would change.
|
|
#
?
Nov 5, 2016 22:31
|
|
|
Alan Smithee posted:I haven't been one to take the troll bait much but just to clarify, does this include quoting a poster you disagree with and attacking their position? You can attack arguments, not people
|
|
#
?
Nov 5, 2016 22:33
|
|
|
cant cook creole bream posted:Yes. It's really not easy to incorporate those at all. The demographic of early voters is not a random sample, so you can't just lump them with the usual polls.
|
|
#
?
Nov 5, 2016 22:35
|
|
|
How do you guys feel about Ohio? https://twitter.com/JaxAlemany/status/795011817840017408
|
|
#
?
Nov 5, 2016 22:37
|
|
|
nachos posted:Why do you even need to intimidate voters in loving Arkansas? e: Also probably a lot of it is to keep "those people" in line. I imagine a lot of the Trump campaign, shall we say, isn't very concerned about black folks in rural Michigan but are VERY concerned about black folks in Arkansas and South Carolina. For some reason. Nessus fucked around with this message at 22:39 on Nov 5, 2016 |
|
#
?
Nov 5, 2016 22:37
|
|
|
nachos posted:Why do you even need to intimidate voters in loving Arkansas? Intimidation is an end in itself.
|
|
#
?
Nov 5, 2016 22:39
|
|
|
nachos posted:Why do you even need to intimidate voters in loving Arkansas? What's the Arkansasian opinion of the Clintons anyway? Obviously, she wont win that state, but I kind of wonder how Bill managed to become governor back in the day. Do they all claim that he stole the election and deceived the voters? I don't really know a lot about local politics of an unimportant state in a foreign country from before my birth. Unless that is too much of a tangent. I don't want to break the rules. cant cook creole bream fucked around with this message at 22:44 on Nov 5, 2016 |
|
#
?
Nov 5, 2016 22:39
|
|
|
cant cook creole bream posted:What's the Arkansasian opinion of the Clintons anyway? Obviously, she wont win that state, but I kind of wonder how Bill managed to become governor back in the day. Do they all claim that he stole the election and deceived the voters? I don't really know a lot about local politics of an unimportant state in a foreign country from before my birth.
|
|
#
?
Nov 5, 2016 22:46
|
|
|
cant cook creole bream posted:What's the Arkansasian opinion of the Clintons anyway? Obviously, she wont win that state, but I kind of wonder how Bill managed to become governor back in the day. Do they all claim that he stole the election and deceived the voters? I don't really know a lot about local politics of an unimportant state in a foreign country from before my birth. I don't know about today's view of him in retrospect, but as gov his earlier career was contentious and later he became quite popular. He lost re-election after his first term primarily for two reasons: he raised taxes (specifically licensing fees) and he got burned hard after the Mariela boatlift (think Scarface) - 300 Cubans out of the thousands kept in Arkansas camps escaped and rampaged leading to many injuries. He was re-elected in the next session and was popular because while he again raised taxes, he also brought in a ton of jobs to Arkansas. They were not high-paying jobs but unemployment went way down. Also he and Hillary worked hard to improve education and create more stringent requirements for school accreditation. I can't find any polls for the period, unfortunately. http://www.encyclopediaofarkansas.net/encyclopedia/entry-detail.aspx?entryID=95
|
|
#
?
Nov 5, 2016 22:52
|
|
|
Just had a canvasser for HRC, a woman who appeared to be in her 70s, walk through my apartment complex knocking on doors to remind people to vote. This is in Arizona. Literally the first time in my entire 37 years that I have actually seen that here. edit: she had a comprehensive list of D voters to contact.
|
|
#
?
Nov 5, 2016 22:52
|
|
|
ReV VAdAUL posted:A lot of the criticism Silver is getting does seem to be because he isn't telling people what they want to hear. Just look at the way people get angry about Hillary's percentage dropping a few points after a round of bad polling - it's pretty obviously not a sober analysis of methodological flaws, it's people becoming upset at a source of anxiety inducing stimulus, just like some people lose their mind every time there's a negative news cycle for Hillary. People want to see good news like good early voting numbers reflected in 538 because they want comfort from the terrifying possibility of a Trump presidency. They got used to using 538 as that comforting authority figure when Clinton was up hugely, and now they're feeling betrayed, they need their fix. No one gives a poo poo about the soundness of Silver's methods, no one gets emotional about minor misapplication of statistics.
|
|
#
?
Nov 5, 2016 22:54
|
|
|
James Garfield posted:It seems very unlikely that Nate Silver is intentionally favoring the Republicans (rather more likely that he is intentionally using a very high uncertainty, but whatever), but it's almost certain that Nate Silver knows how few of his audience realize that his model predicts Clinton winning about half of the light red colored swing states. Maybe Nate is giving some kind of meta commentary on overestimating vs. underestimating Black Swan events.
|
|
#
?
Nov 5, 2016 22:57
|
|
|
http://fivethirtyeight.com/features/election-update-the-campaign-is-almost-over-and-heres-where-we-stand/538 posted:Now that has become more apparent in the polling, and roughly a third of Trump’s 35 percent chance of victory reflects cases where he just barely gets over the hump in the Electoral College despite losing the popular vote. Man, I just don't know. Doesn't there have to be something funky happening behind the scenes to both assume one candidate has a 1 in 3 chance of winning, but that 1 in 3 of the times he wins it's by a popular vote loss and electoral college win? Overall, the model is giving a 10% chance of the election having a popular vote and electoral college winner split.
|
|
#
?
Nov 5, 2016 23:00
|
|
|
Periodiko posted:Just look at the way people get angry about Hillary's percentage dropping a few points after a round of bad polling - it's pretty obviously not a sober analysis of methodological flaws, it's people becoming upset at a source of anxiety inducing stimulus, just like some people lose their mind every time there's a negative news cycle for Hillary. While I mostly agree and admit to feel this way this last part is quite incorrect. As a statistics nerd, I feel it's my obligation and solemn duty to point at every single mathematical flaw and complain about it.  Joking aside, I love chatting about polls. It really feels like I'm learning a lot of stuff, whenever I read this thread. It's really one of the more informative places on the internet. cant cook creole bream fucked around with this message at 23:03 on Nov 5, 2016 |
|
#
?
Nov 5, 2016 23:00
|
|
|
disjoe posted:Maybe Nate is giving some kind of meta commentary on overestimating vs. underestimating Black Swan events. Draw the shape parameter for the half t out of Nassim Taleb's rear end
|
|
#
?
Nov 5, 2016 23:00
|
|
|
Honestly my biggest problem with Nate isn't his maps and stuff. Sure, it's conservative. Whatever. It's the think pieces or whatever about how Trump is totally gonna win you guys! that he seems to constantly loving put out or at least tweet.
|
|
#
?
Nov 5, 2016 23:00
|
|
|
nachos posted:Why do you even need to intimidate voters in loving Arkansas? There were brown people in the room...
|
|
#
?
Nov 5, 2016 23:00
|
|
|
RVProfootballer posted:http://fivethirtyeight.com/features/election-update-the-campaign-is-almost-over-and-heres-where-we-stand/
|
|
#
?
Nov 5, 2016 23:01
|
|
|
Even in 2012, 538 didn't provide any reliable source of comfort. Obama was never more than a 60% favorite for the vast majority of the race. And people panicked, totally lost their poo poo when his odds plummeted after the first Obama-Romney debate, where Romney swept the floor with him. It wasn't until the eleventh hour that it actually broke for Obama.
|
|
#
?
Nov 5, 2016 23:02
|
|
|
berserker posted:Just had a canvasser for HRC, a woman who appeared to be in her 70s, walk through my apartment complex knocking on doors to remind people to vote. This is in Arizona. Literally the first time in my entire 37 years that I have actually seen that here. 
|
|
#
?
Nov 5, 2016 23:03
|
|
|
djw175 posted:Honestly my biggest problem with Nate isn't his maps and stuff. Sure, it's conservative. Whatever. It's the think pieces or whatever about how Trump is totally gonna win you guys! that he seems to constantly loving put out or at least tweet. Everyone with a degree who isn't paid by Trump to say he'll win, says he'll lose.
|
|
#
?
Nov 5, 2016 23:02
|
|
|
cant cook creole bream posted:I'm not saying he does that. but theoretically it would almost double his pageviews. Republicans aren't going to watch one of several sites, which show an uninteresting loss for their team. cant cook creole bream posted:I'm not saying he does that. but theoretically it would almost double his pageviews. Republicans aren't going to watch one of several sites, which show an uninteresting loss for their team. Disney Web Design Budget>>>> Princeton University's Web Design Budget They make good infographs
|
|
#
?
Nov 5, 2016 23:03
|
|
|
Cingulate posted:Trump is underperforming in red states, but his chances at states like Ohio are not as bad. So red states like Texas "eat" some of Clinton's margin. I get the reasoning why it might be that high (and it's actually 12.6% of a popular/electoral split overall), I just don't believe it can actually be that high. Edit: I don't have any good argument really, other than that seems absurdly high for something that very rarely happens. Every election is to some extent its own special snowflake of course, but I can't square that probability with reality.
|
|
#
?
Nov 5, 2016 23:03
|
|
|
Pastrymancy posted:Disney Web Design Budget>>>> Princeton University's Web Design Budget I see their style. I know it. This is my people. ggplot colors, man.
|
|
#
?
Nov 5, 2016 23:06
|
|
|
Cingulate posted:He has literally never done that. You're making stuff up. What he's regularly said is that it's not unlikely that Trump will win, but that it's more likely that Clinton will win, which seems completely fair. Cingulate posted:ggplot colors, man. He doesn't use R though!
|
|
#
?
Nov 5, 2016 23:05
|
|
|
CottonWolf posted:What he's regularly said is that it's not unlikely that Trump will win, but that it's more likely that Clinton will win, which seems completely fair. I may have been exaggerating a bit, but the amount of times he does it and how often he did it even when Clinton was +8, he's harping on it a lot for something even his models say isn't particularly likely.
|
|
#
?
Nov 5, 2016 23:07
|
|
|
djw175 posted:Honestly my biggest problem with Nate isn't his maps and stuff. Sure, it's conservative. Whatever. It's the think pieces or whatever about how Trump is totally gonna win you guys! that he seems to constantly loving put out or at least tweet. A secret plot line of this election is Nate being ultra stressed because 538 is likely getting the ax after the election.
|
|
#
?
Nov 5, 2016 23:06
|
|
|
Actually it is quite "unlikely" that Trump wins.
|
|
#
?
Nov 5, 2016 23:08
|
|
|
|
| # ? Jun 11, 2024 06:47 |
|
|
Nate and crew just really aren't well suited to the "multiple articles every weekday" format they're in now.
|
|
|
#
?
Nov 5, 2016 23:08
|
|