|
Read the rules in the OP before posting, please.In order to ensure that this thread continues to meet TL standards and follows the proper guidelines, we will be enforcing the rules in the OP more strictly. Be sure to give them a re-read to refresh your memory! The vast majority of you are contributing in a healthy way, keep it up! NOTE: When providing a source, explain why you feel it is relevant and what purpose it adds to the discussion if it's not obvious. Also take note that unsubstantiated tweets/posts meant only to rekindle old arguments can result in a mod action. |
holy shit, I agree with biologymajor for once
|
On September 13 2016 06:00 FiWiFaKi wrote:Show nested quote +On September 13 2016 05:58 Plansix wrote:On September 13 2016 05:54 FiWiFaKi wrote: Is there any good election analysis websites?
538 has been absolute garbage. What kind of statistical model would make the "Now Cast" go from 33.2% for Trump on Sept 7th, down to 25.6% on the 9th... And now climb back up.
There is literally nothing that happened that could could represent a roughly 1.5 point loss for Trump over Hillary over that time (not looking at polls, but looking at real life)... 1.5 is a rough figure which comes from the standard deviations they use for their distributions. 538 only uses polls and nothing else. The "polls plus" prediction is only based on trends from the last bunch of elections and how the states preformed and shifted. So the changed was caused by the polls changing. Well then it's model is really poor, and giving way too much weight to one or two polls, that change is way too erratic and nothing close to reality. Their ability to predict the last couple elections with a high level of accuracy says otherwise. From my understanding, very little of their formula has changed. They are very clear why they weight some polls more than others, mostly based on who is polled and if they are “likely voters”.
|
On September 13 2016 05:54 FiWiFaKi wrote: Is there any good election analysis websites?
538 has been absolute garbage. What kind of statistical model would make the "Now Cast" go from 33.2% for Trump on Sept 7th, down to 25.6% on the 9th... And now climb back up.
There is literally nothing that happened that could could represent a roughly 1.5 point loss for Trump over Hillary over that time (not looking at polls, but looking at real life)... 1.5 is a rough figure which comes from estimating vote % from winning chances. The nowcast is fairly pointless, it's basically just what the polls in the last 24 hours say. They should probably get rid of it after this election as it seems just like a gimmick that wasn't meant to be taken seriously in the first place.
The other two models are a lot better.
edit: Also, if you're curious as to why a change happened, they have this page for you that shows the polls they've added in : http://projects.fivethirtyeight.com/2016-election-forecast/updates/
|
On September 13 2016 05:59 biology]major wrote: Optics in the debates might change if HRC fully accepts the sick role and sticks to policy. The usual mud slinging might just enable sympathizers for her due to being unhealthy and old, meaning Trump is going to have to maintain a presidential tone and not get involved in petty attacks. I've been considering this possibility as well. It'll be fantastic if this is able to get spun into a source of sympathy. Not to mention, when she does make a recovery, it will make her look much stronger. Was sick, had pneumonia, still did stuff, pushed herself too far and then got right back into it.
However, this is all assuming she doesn't have anything worse going on
|
On September 13 2016 06:02 Plansix wrote:Show nested quote +On September 13 2016 06:00 FiWiFaKi wrote:On September 13 2016 05:58 Plansix wrote:On September 13 2016 05:54 FiWiFaKi wrote: Is there any good election analysis websites?
538 has been absolute garbage. What kind of statistical model would make the "Now Cast" go from 33.2% for Trump on Sept 7th, down to 25.6% on the 9th... And now climb back up.
There is literally nothing that happened that could could represent a roughly 1.5 point loss for Trump over Hillary over that time (not looking at polls, but looking at real life)... 1.5 is a rough figure which comes from the standard deviations they use for their distributions. 538 only uses polls and nothing else. The "polls plus" prediction is only based on trends from the last bunch of elections and how the states preformed and shifted. So the changed was caused by the polls changing. Well then it's model is really poor, and giving way too much weight to one or two polls, that change is way too erratic and nothing close to reality. Their ability to predict the last couple elections with a high level of accuracy says otherwise. From my understanding, very little of their formula has changed. They are very clear why they weight some polls more than others, mostly based on who is polled and if they are “likely voters”.
Lets be real, nobody needed to look at 538, just polls on their own to be able to predict the 2008 and 2012 election, worst case you have a 50/50 chance.
The reality is the model is not predicting reality in the smaller shifts, even if it's getting the overall shift correctly, that's a pretty big deal. I gave one example of how clearly it treats data incorrectly, in a way that doesn't represent reality.
|
On September 13 2016 05:04 Plansix wrote: I would have to check the laws for MA, but I know that candidates that pass away stay on the ballot in many states. If they somehow win, a special election is held. I’m not sure it works the same for the Oval office, but the convention is normally the point where the party is locked it. Cannot change their minds. There is no “throwing Clinton under the bus so hard she will drop out,” option at this point. that does not sound like a well designed system.... i mean the us is not really short on brains, cultural history and language capabilities.. i would somehow expect them to come up and formulate something better than: "first past the post of those 2* unchangeable options decided months in advance of the election even if one, two or all candidates on the ballot die
*unwinnable token characters may be added to the bottom of the list for decoration"
edit: even the catholic church made it possible for Ratzi do resign even though only god can call him from his post, when you lose a flexiblity of rules contest to the catholic church, you have a problem.
|
On September 13 2016 06:05 FiWiFaKi wrote:Show nested quote +On September 13 2016 06:02 Plansix wrote:On September 13 2016 06:00 FiWiFaKi wrote:On September 13 2016 05:58 Plansix wrote:On September 13 2016 05:54 FiWiFaKi wrote: Is there any good election analysis websites?
538 has been absolute garbage. What kind of statistical model would make the "Now Cast" go from 33.2% for Trump on Sept 7th, down to 25.6% on the 9th... And now climb back up.
There is literally nothing that happened that could could represent a roughly 1.5 point loss for Trump over Hillary over that time (not looking at polls, but looking at real life)... 1.5 is a rough figure which comes from the standard deviations they use for their distributions. 538 only uses polls and nothing else. The "polls plus" prediction is only based on trends from the last bunch of elections and how the states preformed and shifted. So the changed was caused by the polls changing. Well then it's model is really poor, and giving way too much weight to one or two polls, that change is way too erratic and nothing close to reality. Their ability to predict the last couple elections with a high level of accuracy says otherwise. From my understanding, very little of their formula has changed. They are very clear why they weight some polls more than others, mostly based on who is polled and if they are “likely voters”. Lets be real, nobody needed to look at 538, just polls on their own to be able to predict the 2008 and 2012 election, worst case you have a 50/50 chance. The reality is the model is not predicting reality in the smaller shifts, even if it's getting the overall shift correctly, that's a pretty big deal. I gave one example of how clearly it treats data incorrectly, in a way that doesn't represent reality. I think you hit the nail on the head, that predicting the state of the election on a day by day basis with any accuracy is extremely difficult. Especially in this modern era when a close election benefits the news media’s bottom line. 538 is likely the best you are going to get.
|
On September 13 2016 06:02 Nevuk wrote:Show nested quote +On September 13 2016 05:54 FiWiFaKi wrote: Is there any good election analysis websites?
538 has been absolute garbage. What kind of statistical model would make the "Now Cast" go from 33.2% for Trump on Sept 7th, down to 25.6% on the 9th... And now climb back up.
There is literally nothing that happened that could could represent a roughly 1.5 point loss for Trump over Hillary over that time (not looking at polls, but looking at real life)... 1.5 is a rough figure which comes from estimating vote % from winning chances. The nowcast is fairly pointless, it's basically just what the polls in the last 24 hours say. They should probably get rid of it after this election as it seems just like a gimmick that wasn't meant to be taken seriously in the first place. The other two models are a lot better. edit: Also, if you're curious as to why a change happened, they have this page for you that shows the polls they've added in : http://projects.fivethirtyeight.com/2016-election-forecast/updates/
Yeah, I agree, even the polls only forecast is quite poor, polls-plus is the only one that looks reasonable to me.
I suppose the my biggest complaint, but hard to solve... is they give weighting to the polls based on their historic accuracy. I mean sure, it makes sense, but I think polls are so hit and miss with relatively small sample sizes, that often bad polling practices can get reasonably close results. I would much rather they review their methodologies and rate them based on that, or at least have a components that rates them on that.
I'm under the impression that past performance as an indicator of future performance (in terms of polls), is not nearly as effective a quantity compared to say employee performance. That, and of course the polls shouldn't create massive swings, yet still respond quickly to a sudden shift in public eye.
So like when 90% of polls over two days give Trump +5 over what he had before, that should be a sharp rise, but if there's 15 polls which are all over the place (which is what that link showed me), and they have a net +5 for Trump, the rise should be way slower.
|
If you think you can do better fiwi; then by all means. reality shifts in weird ways, and does so quite quickly; hence why the stock market shifts so much all the time. predicting the future is just really really hard.
|
On September 13 2016 06:21 FiWiFaKi wrote:Show nested quote +On September 13 2016 06:02 Nevuk wrote:On September 13 2016 05:54 FiWiFaKi wrote: Is there any good election analysis websites?
538 has been absolute garbage. What kind of statistical model would make the "Now Cast" go from 33.2% for Trump on Sept 7th, down to 25.6% on the 9th... And now climb back up.
There is literally nothing that happened that could could represent a roughly 1.5 point loss for Trump over Hillary over that time (not looking at polls, but looking at real life)... 1.5 is a rough figure which comes from estimating vote % from winning chances. The nowcast is fairly pointless, it's basically just what the polls in the last 24 hours say. They should probably get rid of it after this election as it seems just like a gimmick that wasn't meant to be taken seriously in the first place. The other two models are a lot better. edit: Also, if you're curious as to why a change happened, they have this page for you that shows the polls they've added in : http://projects.fivethirtyeight.com/2016-election-forecast/updates/ Yeah, I agree, even the polls only forecast is quite poor, polls-plus is the only one that looks reasonable to me. I suppose the my biggest complaint, but hard to solve... is they give weighting to the polls based on their historic accuracy. I mean sure, it makes sense, but I think polls are so hit and miss with relatively small sample sizes, that often bad polling practices can get reasonably close results. I would much rather they review their methodologies and rate them based on that, or at least have a components that rates them on that. I'm under the impression that past performance as an indicator of future performance (in terms of polls), is not nearly as effective a quantity compared to say employee performance. That, and of course the polls shouldn't create massive swings, yet still respond quickly to a sudden shift in public eye. So like when 90% of polls over two days give Trump +5 over what he had before, that should be a sharp rise, but if there's 15 polls which are all over the place (which is what that link showed me), and they have a net +5 for Trump, the rise should be way slower.
Their ratings are based in part on methodology, actually :
FiveThirtyEight’s pollster ratings are calculated by analyzing the historical accuracy and the methodology of each firm’s polls. Accuracy scores account for the type of election, a poll’s sample size, the performance of other polls surveying the same race and other factors. We also calculate measures of statistical bias in the polls.
Their full article is here: http://fivethirtyeight.com/features/how-fivethirtyeight-calculates-pollster-ratings/ (though that snippet is from their rankings list specifically)
Employee talent is a harder thing to measure, as they don't go into specifics on why some pollsters have such bad ratings, but it is something they've stated as being one of the reasons for some of their highest ratings (specifically Ann Selzer's Iowa polls), and Ziegler did have that hilarious interview with Nate Silver in 2008 on "push polls" which I imagined colored perceptions some : http://fivethirtyeight.com/features/interview-with-john-ziegler-on-zogby/ .
Choice quote :
NS: Thank you, have a good day. JZ: Go fuck yourself.
When Ipsos changed methodology earlier this year they went from favoring Clinton fairly heavily to having essentially no bias, and 538 adjusted the house effect based on this change (and were a little critical of changing methodology halfway through an election season).
538 puts a lot of thought into their models, but their complexity frequently leaves them writing walls of text when the short snippet descriptions would do a much better job of explaining it.
|
One of the things about 538 is that they weight polls based on past performance and the methodology of polling. Nate Silver talks about inaccurate or outlier polls a lot and how those results are reached. He doesn’t outright stay they are “rigged”, but points out that a news network might switch up their methodology if the race isn’t close.
I think it is more informative to read the articles about the polling, rather than stare at the data itself to find meaning.
Edit: Nevek crushes it. And that interview quote is hilarious.
|
On September 13 2016 06:28 zlefin wrote: If you think you can do better fiwi; then by all means. reality shifts in weird ways, and does so quite quickly; hence why the stock market shifts so much all the time. predicting the future is just really really hard.
That's subjective, looking at NYSE over the last 2 years, it's been bound at +/-10% with the exception of a couple days. Either way, I don't think it's really comparable, since the stock market is speculative trading, while come on... This is the election.
Most polls show around 80% of the people won't change their mind now, just think of yourself... How much would it take for you to support Trump if you're a Hillary supporter? I few days of quiet won't make these changes.
Of course I'm not going to make a website to interpret these polls, but I do find that I get a better oversight by looking at the national polls and reading the papers/articles associated with them, opposed to 538. This doesn't give the state perspective, which of course would be very time consuming to do on its own.
|
On September 13 2016 05:29 Acrofales wrote:Show nested quote +On September 13 2016 05:07 Danglars wrote:On September 13 2016 01:49 Dan HH wrote:On September 13 2016 01:43 Danglars wrote:On September 13 2016 00:56 xDaunt wrote:I'll just let this guy make the point for me:Hillary Clinton's presidential campaign just made a massive error. We'll know within the next few weeks if the error will prove to be catastrophic.
On Sunday, Clinton abruptly left a Manhattan ceremony marking the 15th anniversary of the 9/11 terrorist attacks. A video shows her shakily stumbling while trying to get into a van to leave. The candidate's physician later offered this explanation: Clinton has had an allergy-related cough for some time, and during an examination on Friday, the Democratic nominee was diagnosed with pneumonia, put on antibiotics, and told to take time out to rest. She became overheated and dehydrated during Sunday morning's event, which led her to collapse. She's now home in Chappaqua and on the road to recovery.
Compare this timeline to details from Hillary Clinton's public schedule and behavior over these same two days.
After Clinton was diagnosed with pneumonia and put on antibiotics, she did not, as her physician recommended, take time out to rest. Instead, she attended a fundraiser featuring Barbra Streisand. Then on Sunday morning, she attended the 9/11 commemoration, became "overheated," and woozily wobbled rather dramatically. Ninety minutes later she exited her daughter Chelsea's apartment building to tell the press she was "feeling great." The Secret Service permitted a young girl to come over to give the candidate a hug.
It was only a few hours later when her campaign finally announced that she has pneumonia and is recovering.
The most charitable reading of this timeline is that her campaign — presumably with the blessing and perhaps insistence of the candidate — fully intended to keep her illness a secret from the public. Let's be clear about what this means: Her campaign intended to lie. Even though doing so would require her to keep up a public schedule that might well make her condition worse and require ever-more elaborate forms of concealment. Because, of course, to curtail her schedule would raise questions that might reveal the truth.
So even after she collapsed, the campaign decided the ruse would continue. It arranged for the candidate to make her curbside declaration of wellness, even bringing on the girl to give her a "spontaneous" hug. (Clinton's protection detail would never have permitted a genuinely spontaneous embrace on the street, even by a child.)
It's easy to understand why the Clinton campaign would want to keep this kind of news a secret. The candidate doesn't trust the media. The right has been hitting her over supposed health issues for months (and even years), and the assault has picked up in intensity over the past week or so — since Clinton found herself in the midst of an extended coughing fit at a campaign event in Cleveland. Then there's the gender dynamic. Donald Trump presents himself as a hyper-masculine tough guy, while Clinton is the first female presidential nominee. The Clinton camp is probably twice as terrified of their candidate looking frail as a less path-breaking campaign would be.
So the campaign chose to lie. The potential reward was considerable: namely, an absence of politically damaging news stories about Clinton's medical condition. But the risk was enormous — and it's blown up in their faces. Because now the story isn't just that Clinton is ill. It's that, once again, she's untrustworthy — and this time about her own health.
That's why the announcement that she has pneumonia will only fuel more speculation about Clinton's physical condition, with potentially no end in sight. The world saw her collapse, and 90 minutes later, the candidate looked America in the eye and proclaimed that she was feeling great. Except now we know that she wasn't.
Not long after this charade, someone on the campaign staff made the call to come clean. But it may well have been too late.
The best the campaign can hope for now is that Clinton recovers quickly and soon looks healthy in her public appearances. Then maybe the topic will recede into the background of the campaign. The candidate got sick, but then she got better. End of story.
But if she doesn't recover quickly? If she appears weak and frail for more than a few days? Then, yes, she'll face perfectly reasonable questions about whether she's physically up to serving as president. But worse, she'll confront lingering doubts about what, precisely, is ailing her. "It's pneumonia," the campaign will proclaim over and over again. To which a skeptical America will justifiably reply, "Yes, we can tell that you'd like us to think so. But we have no reason to trust that's true."
Political trust is a fragile thing. Once it's gone, it's exceedingly difficult to get back — and without it, there's no basis on which to dismiss conspiracy theories that even normally level-headed observers will begin, for perfectly understandable reasons, to entertain.
Like so many of the scandals and pseudo-scandals that have dogged Hillary Clinton and her husband through the years, this one needs to be recognized as entirely self-inflicted. The campaign now has to live with the consequences of having chosen to lie to get out of a problem. You Hillary supporters need to shake yourselves out of your delusions. Her campaign clearly messed up, and I'm not sure why it's so hard for y'all to see it and concede that point (edit: Actually, I do know why, but I'm going to be nice). It's not that she lies, she just needs to hire on better liars. This issue comes on the heels of a related topic: it's not that she's corrupt, she's incompetent at hiding her corruption. I see Hunts abandoned asking xDaunt to elaborate about Clinton's clear corruption after the nth time of no answer. But if you say that she is incompetent at hiding her corruption, I do hope at least you have some examples. You'll have to buy a book if you want all her corruption spelled out. The reduced form comes down to trying and failing to hide a private server to conduct business free from FOIA requests, with a pattern of lying when questions arose about classified emails and deleting emails even while they were subpoena'd. + Show Spoiler [hilarious as hell "oh shit"] +She also met with and assisted foreign national donors to her clinton family foundation while working as secretary of state. The incompetence story is huge and shows no signs of stopping, though it's hard to get better than the current admissions out there. She instructed aides to remove the classified headers from documents before sending them to her, and later claimed she couldn't identify other emails as classified because there were no headers. Her foundation accidentally didn't disclose tens of millions of dollars of donations from foreign governments for three years, and had to refile. The deletions were justified because they were just joga emails etc, even though the latest 15,000 contained at least one benghazi email. She claimed Comey corroborated her "truthful" statements made with respect to the investigation and what she told the American people. I've been reading the responses in this thread and the acrobatics, and my only takeaway is that paying for government favors shouldn't surprise anybody, and it was only natural for Hillary to cover up her violations of the espionage act because nobody should care about classifications regardless. To say that another way since the thread's been heading there lately, everybody knows Clinton is corrupt and has used her office to commit crimes without consequence and enrich herself. (shitposter clarification: she has been dragged to interviews & hearings, so it wasn't entirely without consequence) Paying for government favors? You really think she'd stoop to Trump levels of corruption? As for being dragged to hearings, that's would be the partisan politics at work. More interesting is that none of the investigations those hearings were part of found any actual wrongdoing... Do I like her personally? No. And I am certainly not a fan of the back chamber deal making that is the norm in Washington, which Hilary is obviously both a part of and adept at. But actual corruption goes a lot further than that. Anybody who actually thinks that it would take her stooping to arrive at Trump levels of corruption slept through the nineties. If you read what I wrote and it's "back chamber deals" not rising to "actual corruption, we're probably done here.
On September 13 2016 05:55 Dan HH wrote:Show nested quote +On September 13 2016 05:07 Danglars wrote:On September 13 2016 01:49 Dan HH wrote:On September 13 2016 01:43 Danglars wrote:On September 13 2016 00:56 xDaunt wrote:I'll just let this guy make the point for me:Hillary Clinton's presidential campaign just made a massive error. We'll know within the next few weeks if the error will prove to be catastrophic.
On Sunday, Clinton abruptly left a Manhattan ceremony marking the 15th anniversary of the 9/11 terrorist attacks. A video shows her shakily stumbling while trying to get into a van to leave. The candidate's physician later offered this explanation: Clinton has had an allergy-related cough for some time, and during an examination on Friday, the Democratic nominee was diagnosed with pneumonia, put on antibiotics, and told to take time out to rest. She became overheated and dehydrated during Sunday morning's event, which led her to collapse. She's now home in Chappaqua and on the road to recovery.
Compare this timeline to details from Hillary Clinton's public schedule and behavior over these same two days.
After Clinton was diagnosed with pneumonia and put on antibiotics, she did not, as her physician recommended, take time out to rest. Instead, she attended a fundraiser featuring Barbra Streisand. Then on Sunday morning, she attended the 9/11 commemoration, became "overheated," and woozily wobbled rather dramatically. Ninety minutes later she exited her daughter Chelsea's apartment building to tell the press she was "feeling great." The Secret Service permitted a young girl to come over to give the candidate a hug.
It was only a few hours later when her campaign finally announced that she has pneumonia and is recovering.
The most charitable reading of this timeline is that her campaign — presumably with the blessing and perhaps insistence of the candidate — fully intended to keep her illness a secret from the public. Let's be clear about what this means: Her campaign intended to lie. Even though doing so would require her to keep up a public schedule that might well make her condition worse and require ever-more elaborate forms of concealment. Because, of course, to curtail her schedule would raise questions that might reveal the truth.
So even after she collapsed, the campaign decided the ruse would continue. It arranged for the candidate to make her curbside declaration of wellness, even bringing on the girl to give her a "spontaneous" hug. (Clinton's protection detail would never have permitted a genuinely spontaneous embrace on the street, even by a child.)
It's easy to understand why the Clinton campaign would want to keep this kind of news a secret. The candidate doesn't trust the media. The right has been hitting her over supposed health issues for months (and even years), and the assault has picked up in intensity over the past week or so — since Clinton found herself in the midst of an extended coughing fit at a campaign event in Cleveland. Then there's the gender dynamic. Donald Trump presents himself as a hyper-masculine tough guy, while Clinton is the first female presidential nominee. The Clinton camp is probably twice as terrified of their candidate looking frail as a less path-breaking campaign would be.
So the campaign chose to lie. The potential reward was considerable: namely, an absence of politically damaging news stories about Clinton's medical condition. But the risk was enormous — and it's blown up in their faces. Because now the story isn't just that Clinton is ill. It's that, once again, she's untrustworthy — and this time about her own health.
That's why the announcement that she has pneumonia will only fuel more speculation about Clinton's physical condition, with potentially no end in sight. The world saw her collapse, and 90 minutes later, the candidate looked America in the eye and proclaimed that she was feeling great. Except now we know that she wasn't.
Not long after this charade, someone on the campaign staff made the call to come clean. But it may well have been too late.
The best the campaign can hope for now is that Clinton recovers quickly and soon looks healthy in her public appearances. Then maybe the topic will recede into the background of the campaign. The candidate got sick, but then she got better. End of story.
But if she doesn't recover quickly? If she appears weak and frail for more than a few days? Then, yes, she'll face perfectly reasonable questions about whether she's physically up to serving as president. But worse, she'll confront lingering doubts about what, precisely, is ailing her. "It's pneumonia," the campaign will proclaim over and over again. To which a skeptical America will justifiably reply, "Yes, we can tell that you'd like us to think so. But we have no reason to trust that's true."
Political trust is a fragile thing. Once it's gone, it's exceedingly difficult to get back — and without it, there's no basis on which to dismiss conspiracy theories that even normally level-headed observers will begin, for perfectly understandable reasons, to entertain.
Like so many of the scandals and pseudo-scandals that have dogged Hillary Clinton and her husband through the years, this one needs to be recognized as entirely self-inflicted. The campaign now has to live with the consequences of having chosen to lie to get out of a problem. You Hillary supporters need to shake yourselves out of your delusions. Her campaign clearly messed up, and I'm not sure why it's so hard for y'all to see it and concede that point (edit: Actually, I do know why, but I'm going to be nice). It's not that she lies, she just needs to hire on better liars. This issue comes on the heels of a related topic: it's not that she's corrupt, she's incompetent at hiding her corruption. I see Hunts abandoned asking xDaunt to elaborate about Clinton's clear corruption after the nth time of no answer. But if you say that she is incompetent at hiding her corruption, I do hope at least you have some examples. You'll have to buy a book if you want all her corruption spelled out. The reduced form comes down to trying and failing to hide a private server to conduct business free from FOIA requests, with a pattern of lying when questions arose about classified emails and deleting emails even while they were subpoena'd. + Show Spoiler [hilarious as hell "oh shit"] +She also met with and assisted foreign national donors to her clinton family foundation while working as secretary of state. The incompetence story is huge and shows no signs of stopping, though it's hard to get better than the current admissions out there. She instructed aides to remove the classified headers from documents before sending them to her, and later claimed she couldn't identify other emails as classified because there were no headers. Her foundation accidentally didn't disclose tens of millions of dollars of donations from foreign governments for three years, and had to refile. The deletions were justified because they were just yoga emails etc, even though the latest 15,000 contained at least one benghazi email. She claimed Comey corroborated her "truthful" statements made with respect to the investigation and what she told the American people. I've been reading the responses in this thread and the acrobatics, and my only takeaway is that paying for government favors shouldn't surprise anybody, and it was only natural for Hillary to cover up her violations of the espionage act because nobody should care about classifications regardless. To say that another way since the thread's been heading there lately, everybody knows Clinton is corrupt and has used her office to commit crimes without consequence and enrich herself. (shitposter clarification: she has been dragged to interviews & hearings, so it wasn't entirely without consequence) I was hoping for something along the lines of person puts money in then receives political favor that can be reasonably correlated with it. When you say that not only is she corrupt (something that I'm very willing to consider), but she is incompetent at hiding said corruption, I don't think that 'everyone know she is corrupt', 'she met with donors' and 'she deleted emails' is good enough. When you claim it's so obvious, you need at least something half as clear as the several instances of Trump paying for favors such as with Pam Bondi or Andrew Stein. That would all depend on your level of reasonability. But I just heard the case that everybody knows that Trump is racist and something like half his supporters are in a basket of deplorables. I'm feeling very emboldened to say everybody knows Clinton is corrupt, has used her office to financial profit and to increase her power, has broken laws, and lied about both. I don't really know how clear clear is to you, but I have this nagging feeling that your mirror is broken for looking hard at the failings of people you mostly agree with.
|
On September 13 2016 06:40 FiWiFaKi wrote:Show nested quote +On September 13 2016 06:28 zlefin wrote: If you think you can do better fiwi; then by all means. reality shifts in weird ways, and does so quite quickly; hence why the stock market shifts so much all the time. predicting the future is just really really hard. That's subjective, looking at NYSE over the last 2 years, it's been bound at +/-10% with the exception of a couple days. Either way, I don't think it's really comparable, since the stock market is speculative trading, while come on... This is the election. Most polls show around 80% of the people won't change their mind now, just think of yourself... How much would it take for you to support Trump if you're a Hillary supporter? I few days of quiet won't make these changes. Of course I'm not going to make a website to interpret these polls, but I do find that I get a better oversight by looking at the national polls and reading the papers/articles associated with them, opposed to 538. This doesn't give the state perspective, which of course would be very time consuming to do on its own. I'm talking about how the stock market can go +/- 1% in a single day. sure they average out over time; but those are some awfully big fluctuations on a daily basis. If that can happen; it doesn't seem so surprising that some people are wavering in their decisions, and all sorts of things, including random sampling effects and errors, can shift things around.
|
On September 13 2016 06:37 Plansix wrote: One of the things about 538 is that they weight polls based on past performance and the methodology of polling. Nate Silver talks about inaccurate or outlier polls a lot and how those results are reached. He doesn’t outright stay they are “rigged”, but points out that a news network might switch up their methodology if the race isn’t close.
I think it is more informative to read the articles about the polling, rather than stare at the data itself to find meaning.
The models themselves are pretty high quality and useful if you merely want a snapshot of the current state of the race, but yes, the analysis of the polls is probably the best part of the website.
They've referenced a few times recently that some Democratic readers feel betrayed by them for not giving the reassurance that they did in 2012, where almost every poll consistently showed a close Obama lead. Those readers are missing the point - 2016 polls do not in any way show a race that is reminiscent of 2012. It instead shows a very uncertain race, in their view (that's the biggest difference between them and other pollster aggregates this year - they think it's a less certain race overall, greater chance of a Clinton landslide OR a Trump win than other aggregators). They don't exist to comfort their readers, they exist to describe the state of the race as described via polling (which is why they feel they missed especially badly on Trump in the primaries - the polls always showed him to be a serious candidate , and they ignored it until far too late).
|
United Kingdom13775 Posts
On September 13 2016 06:02 Plansix wrote:Show nested quote +On September 13 2016 06:00 FiWiFaKi wrote:On September 13 2016 05:58 Plansix wrote:On September 13 2016 05:54 FiWiFaKi wrote: Is there any good election analysis websites?
538 has been absolute garbage. What kind of statistical model would make the "Now Cast" go from 33.2% for Trump on Sept 7th, down to 25.6% on the 9th... And now climb back up.
There is literally nothing that happened that could could represent a roughly 1.5 point loss for Trump over Hillary over that time (not looking at polls, but looking at real life)... 1.5 is a rough figure which comes from the standard deviations they use for their distributions. 538 only uses polls and nothing else. The "polls plus" prediction is only based on trends from the last bunch of elections and how the states preformed and shifted. So the changed was caused by the polls changing. Well then it's model is really poor, and giving way too much weight to one or two polls, that change is way too erratic and nothing close to reality. Their ability to predict the last couple elections with a high level of accuracy says otherwise. From my understanding, very little of their formula has changed. They are very clear why they weight some polls more than others, mostly based on who is polled and if they are “likely voters”. What's fucked is the core assumptions that allow them to predict stuff. Some reading: US Survey Research The Endorsement Primary
The first is a comprehensive description of polling methodology from one of the more reliable pollsters (Pew), which admits that in recent times, polling has become increasingly theoretically unsound. They talk about how, for some reason, despite the fact that the sample of people who will answer phone polls is not representative, the results tend to be accurate. Online poll methodologies are also rather unreliable. They also mention how there is no guarantee of future polling accuracy which sort of has manifested itself very strongly in this election.
The second is a description of 538's "Endorsement Primary" model of voting, one of their major fundamental factors based upon the "The Party Decides" theory of voting (the party chooses their candidate, then the electorate usually falls in line). That theory has mostly held for Hillary Clinton (but not as much as you would think; otherwise she would have like 80% of the vote), but it's been a straight-up miss for explaining the rise of Trump whose candidacy is pretty unprecedented. Towards the general election, their theory similarly fails to explain what happens when both candidates are very much hated by the electorate and it's a race to the bottom more than anything else; the theory would also suggest that Hillary Clinton, with her remarkable establishment backing, should have this in a landslide. So the theory doesn't apply and so the models are somewhat fucked because historical data isn't what it used to be.
I think Donald Trump will be quite a topic of interest for political theorists for the decades to come.
|
On September 13 2016 06:35 Nevuk wrote:Show nested quote +On September 13 2016 06:21 FiWiFaKi wrote:On September 13 2016 06:02 Nevuk wrote:On September 13 2016 05:54 FiWiFaKi wrote: Is there any good election analysis websites?
538 has been absolute garbage. What kind of statistical model would make the "Now Cast" go from 33.2% for Trump on Sept 7th, down to 25.6% on the 9th... And now climb back up.
There is literally nothing that happened that could could represent a roughly 1.5 point loss for Trump over Hillary over that time (not looking at polls, but looking at real life)... 1.5 is a rough figure which comes from estimating vote % from winning chances. The nowcast is fairly pointless, it's basically just what the polls in the last 24 hours say. They should probably get rid of it after this election as it seems just like a gimmick that wasn't meant to be taken seriously in the first place. The other two models are a lot better. edit: Also, if you're curious as to why a change happened, they have this page for you that shows the polls they've added in : http://projects.fivethirtyeight.com/2016-election-forecast/updates/ Yeah, I agree, even the polls only forecast is quite poor, polls-plus is the only one that looks reasonable to me. I suppose the my biggest complaint, but hard to solve... is they give weighting to the polls based on their historic accuracy. I mean sure, it makes sense, but I think polls are so hit and miss with relatively small sample sizes, that often bad polling practices can get reasonably close results. I would much rather they review their methodologies and rate them based on that, or at least have a components that rates them on that. I'm under the impression that past performance as an indicator of future performance (in terms of polls), is not nearly as effective a quantity compared to say employee performance. That, and of course the polls shouldn't create massive swings, yet still respond quickly to a sudden shift in public eye. So like when 90% of polls over two days give Trump +5 over what he had before, that should be a sharp rise, but if there's 15 polls which are all over the place (which is what that link showed me), and they have a net +5 for Trump, the rise should be way slower. Their ratings are based in part on methodology, actually : Show nested quote +FiveThirtyEight’s pollster ratings are calculated by analyzing the historical accuracy and the methodology of each firm’s polls. Accuracy scores account for the type of election, a poll’s sample size, the performance of other polls surveying the same race and other factors. We also calculate measures of statistical bias in the polls.
Their full article is here: http://fivethirtyeight.com/features/how-fivethirtyeight-calculates-pollster-ratings/ (though that snippet is from their rankings list specifically) Employee talent is a harder thing to measure, as they don't go into specifics on why some pollsters have such bad ratings, but it is something they've stated as being one of the reasons for some of their highest ratings (specifically Ann Selzer's Iowa polls), and Ziegler did have that hilarious interview with Nate Silver in 2008 on "push polls" which I imagined colored perceptions some : http://fivethirtyeight.com/features/interview-with-john-ziegler-on-zogby/ . Choice quote : When Ipsos changed methodology earlier this year they went from favoring Clinton fairly heavily to having essentially no bias, and 538 adjusted the house effect based on this change (and were a little critical of changing methodology halfway through an election season). 538 puts a lot of thought into their models, but their complexity frequently leaves them writing walls of text when the short snippet descriptions would do a much better job of explaining it.
I actually have read that before, and yes, my apologies, they did do something where they cut the weight of some in half depending on the methodology, but it's still component. They try to fix a poll by saying the bias will be roughly the same each time, and be done with it.... Wtf, that's like saying when I build buildings, I always make it 20% too weak, so I'll just add 20% to my calculations, it's such a gimmick, with not good scientific basis. Just throw those polls out.... They try to claim that if they had error before, they will have error now, and they back that up by having an R value of 0.6, which is a moderate-high correlation, so sure, sounds reasonable, right?
Well no, it's more like saying bad polls are bad and good polls are good. They incorrectly assume that the average error corresponds to a lack of accuracy, when in reality it's more likely that it's a lack of both.
The poll herding mentioned is another silly thing, where individual polls get more accurate the more they are by as much as 2%, because they fudge their data to look better, more the reason to discard them completely.
Anyway, the poll plus forecast on 538 is the best there is on the internet, my main complaint was with regards to Now Cast, and to an extent Polls Only, which don't reflect reality well.
On September 13 2016 06:44 zlefin wrote:Show nested quote +On September 13 2016 06:40 FiWiFaKi wrote:On September 13 2016 06:28 zlefin wrote: If you think you can do better fiwi; then by all means. reality shifts in weird ways, and does so quite quickly; hence why the stock market shifts so much all the time. predicting the future is just really really hard. That's subjective, looking at NYSE over the last 2 years, it's been bound at +/-10% with the exception of a couple days. Either way, I don't think it's really comparable, since the stock market is speculative trading, while come on... This is the election. Most polls show around 80% of the people won't change their mind now, just think of yourself... How much would it take for you to support Trump if you're a Hillary supporter? I few days of quiet won't make these changes. Of course I'm not going to make a website to interpret these polls, but I do find that I get a better oversight by looking at the national polls and reading the papers/articles associated with them, opposed to 538. This doesn't give the state perspective, which of course would be very time consuming to do on its own. I'm talking about how the stock market can go +/- 1% in a single day. sure they average out over time; but those are some awfully big fluctuations on a daily basis. If that can happen; it doesn't seem so surprising that some people are wavering in their decisions, and all sorts of things, including random sampling effects and errors, can shift things around.
Your second part of the post is my exact argument. They are sampling errors, which are just noise, and the Now Cast and Polls Only model is not filtering them out adequately.
|
On September 13 2016 06:45 LegalLord wrote:Show nested quote +On September 13 2016 06:02 Plansix wrote:On September 13 2016 06:00 FiWiFaKi wrote:On September 13 2016 05:58 Plansix wrote:On September 13 2016 05:54 FiWiFaKi wrote: Is there any good election analysis websites?
538 has been absolute garbage. What kind of statistical model would make the "Now Cast" go from 33.2% for Trump on Sept 7th, down to 25.6% on the 9th... And now climb back up.
There is literally nothing that happened that could could represent a roughly 1.5 point loss for Trump over Hillary over that time (not looking at polls, but looking at real life)... 1.5 is a rough figure which comes from the standard deviations they use for their distributions. 538 only uses polls and nothing else. The "polls plus" prediction is only based on trends from the last bunch of elections and how the states preformed and shifted. So the changed was caused by the polls changing. Well then it's model is really poor, and giving way too much weight to one or two polls, that change is way too erratic and nothing close to reality. Their ability to predict the last couple elections with a high level of accuracy says otherwise. From my understanding, very little of their formula has changed. They are very clear why they weight some polls more than others, mostly based on who is polled and if they are “likely voters”. What's fucked is the core assumptions that allow them to predict stuff. Some reading: US Survey ResearchThe Endorsement PrimaryThe first is a comprehensive description of polling methodology from one of the more reliable pollsters (Pew), which admits that in recent times, polling has become increasingly theoretically unsound. They talk about how, for some reason, despite the fact that the sample of people who will answer phone polls is not representative, the results tend to be accurate. Online poll methodologies are also rather unreliable. They also mention how there is no guarantee of future polling accuracy which sort of has manifested itself very strongly in this election. The second is a description of 538's "Endorsement Primary" model of voting, one of their major fundamental factors based upon the "The Party Decides" theory of voting (the party chooses their candidate, then the electorate usually falls in line). That theory has mostly held for Hillary Clinton (but not as much as you would think; otherwise she would have like 80% of the vote), but it's been a straight-up miss for explaining the rise of Trump whose candidacy is pretty unprecedented. Towards the general election, their theory similarly fails to explain what happens when both candidates are very much hated by the electorate and it's a race to the bottom more than anything else; the theory would also suggest that Hillary Clinton, with her remarkable establishment backing, should have this in a landslide. So the theory doesn't apply and so the models are somewhat fucked because historical data isn't what it used to be. I think Donald Trump will be quite a topic of interest for political theorists for the decades to come. My favorite 538 quote on this topic is this:
Though I will admit I'm not entirely sure how unlikely this style of election will be in the future - the internet is a hell of a tool. On the other hand, I kind of think if you went and asked Plato or Aristotle if information should be freely accessible to everyone, they would be horrified. On the other other hand, Plato's philosophy is kind of dickish.
|
I have my own theories on the rise of populism and Trump, but those lack any data to back them up. Needless to say, this is not the last we have heard of this populism. I just hope we can move away from it, because it has a strong contempt for education, experience and reality.
On September 13 2016 06:54 Nevuk wrote:Show nested quote +On September 13 2016 06:45 LegalLord wrote:On September 13 2016 06:02 Plansix wrote:On September 13 2016 06:00 FiWiFaKi wrote:On September 13 2016 05:58 Plansix wrote:On September 13 2016 05:54 FiWiFaKi wrote: Is there any good election analysis websites?
538 has been absolute garbage. What kind of statistical model would make the "Now Cast" go from 33.2% for Trump on Sept 7th, down to 25.6% on the 9th... And now climb back up.
There is literally nothing that happened that could could represent a roughly 1.5 point loss for Trump over Hillary over that time (not looking at polls, but looking at real life)... 1.5 is a rough figure which comes from the standard deviations they use for their distributions. 538 only uses polls and nothing else. The "polls plus" prediction is only based on trends from the last bunch of elections and how the states preformed and shifted. So the changed was caused by the polls changing. Well then it's model is really poor, and giving way too much weight to one or two polls, that change is way too erratic and nothing close to reality. Their ability to predict the last couple elections with a high level of accuracy says otherwise. From my understanding, very little of their formula has changed. They are very clear why they weight some polls more than others, mostly based on who is polled and if they are “likely voters”. What's fucked is the core assumptions that allow them to predict stuff. Some reading: US Survey ResearchThe Endorsement PrimaryThe first is a comprehensive description of polling methodology from one of the more reliable pollsters (Pew), which admits that in recent times, polling has become increasingly theoretically unsound. They talk about how, for some reason, despite the fact that the sample of people who will answer phone polls is not representative, the results tend to be accurate. Online poll methodologies are also rather unreliable. They also mention how there is no guarantee of future polling accuracy which sort of has manifested itself very strongly in this election. The second is a description of 538's "Endorsement Primary" model of voting, one of their major fundamental factors based upon the "The Party Decides" theory of voting (the party chooses their candidate, then the electorate usually falls in line). That theory has mostly held for Hillary Clinton (but not as much as you would think; otherwise she would have like 80% of the vote), but it's been a straight-up miss for explaining the rise of Trump whose candidacy is pretty unprecedented. Towards the general election, their theory similarly fails to explain what happens when both candidates are very much hated by the electorate and it's a race to the bottom more than anything else; the theory would also suggest that Hillary Clinton, with her remarkable establishment backing, should have this in a landslide. So the theory doesn't apply and so the models are somewhat fucked because historical data isn't what it used to be. I think Donald Trump will be quite a topic of interest for political theorists for the decades to come. My favorite 538 quote on this topic is this: https://twitter.com/NateSilver538/status/703714652098846720Though I will admit I'm not entirely sure how unlikely this style of election will be in the future - the internet is a hell of a tool. On the other hand, I kind of think if you went and asked Plato or Aristotle if information should be freely accessible to everyone, they would be horrified. On the other other hand, Plato's philosophy is kind of dickish.
I have deep reservations about current state of the internet and the data coming out that it is making people more polarized and less likely to believe scientific data. On the other hand, that could just be me slowly entering the old man stage of my life that I have been threatening to enter since my mid 20s.
|
United Kingdom13775 Posts
Data means jack shit without useful and accurate assumptions. But plenty of people will make terrible assumptions and claim that "science" backs them up. Needless to say it's complete horseshit.
|
|
|
|