|
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. |
On March 09 2016 13:31 darthfoley wrote:
Just FYI this is also funny because Michigan state schools are on spring break this week, which may have even depressed some Sanders support tonight lol
With how much he crushes young demographics (and college kids being lazy and not preparing for it) that could be a fairly significant chunk.
|
Look at this fucking picture of Rubio that Drudge posted, haha:
![[image loading]](https://pbs.twimg.com/media/CdE27dkUUAAmAxF.jpg)
|
United States22883 Posts
On March 09 2016 13:33 ZeaL. wrote:Show nested quote +On March 09 2016 13:28 TheTenthDoc wrote:On March 09 2016 13:23 ZeaL. wrote:On March 09 2016 13:20 Jibba wrote:On March 09 2016 13:16 ZeaL. wrote:On March 09 2016 13:12 GreenHorizons wrote:On March 09 2016 13:05 ZeaL. wrote:Why is Gladwin County 780-74 for Sanders with 17/18 precincts reporting? Edit: Tweet from 538 earlier today: + Show Spoiler + 538 totally air-balled MI, no if ands or buts about it. They didn't even skim the net let alone touch the rim. I mean they were totally off but it's not completely their fault. Polling for MI has consistently had Sanders down by 10-20% within the last week even. They can't do anything without accurate polls. Even if Hillary pulls it back and somehow wins, this result still shows the polling was way off. They can adjust based on the quality of polling data. Silver did it before, and it's why he ran counter to mainstream polls. They can, but if every single poll is saying Clinton by 10-20% then no matter how you weight the data you're not going to get an expected Sanders win. Edit: I guess if they know the polling data is poor they should have broad confidence intervals. For some reason they neglect to include them in their model. Their model already has confidence intervals. The uncertainty around the point estimate is reflected in the sizes of each candidate's "hump" and, in a two-person race, it's pretty trivial to use this to estimate probabilities. There's virtually no quantitative data in the universe outside of possibly Sanders' campaign office that could predict this outcome. People shouldn't shit on people for not predicting it after the fact. Yeah thinking about it a bit more I understand why the output is presented the way it is. And just for reference this is the data they were working with: ![[image loading]](http://i.imgur.com/3VNWwh0.png) If anything this is an indication of the poor quality of polling companies.
They've always been poor quality polling. We all understand garbage in, garbage out. Nate's magic last time was the system for rating polling quality. The actual vote model probably wasn't that much different than MIT's model.
|
Still not good for Bernie though that he still overall lost on delegates gained since he got whomped in MS. Still though the upset could give him some momentum I guess.
|
On March 09 2016 13:32 farvacola wrote:Show nested quote +On March 09 2016 13:28 TheTenthDoc wrote:On March 09 2016 13:23 ZeaL. wrote:On March 09 2016 13:20 Jibba wrote:On March 09 2016 13:16 ZeaL. wrote:On March 09 2016 13:12 GreenHorizons wrote:On March 09 2016 13:05 ZeaL. wrote:Why is Gladwin County 780-74 for Sanders with 17/18 precincts reporting? Edit: Tweet from 538 earlier today: + Show Spoiler + 538 totally air-balled MI, no if ands or buts about it. They didn't even skim the net let alone touch the rim. I mean they were totally off but it's not completely their fault. Polling for MI has consistently had Sanders down by 10-20% within the last week even. They can't do anything without accurate polls. Even if Hillary pulls it back and somehow wins, this result still shows the polling was way off. They can adjust based on the quality of polling data. Silver did it before, and it's why he ran counter to mainstream polls. They can, but if every single poll is saying Clinton by 10-20% then no matter how you weight the data you're not going to get an expected Sanders win. Edit: I guess if they know the polling data is poor they should have broad confidence intervals. For some reason they neglect to include them in their model. Their model already has confidence intervals. The uncertainty around the point estimate is reflected in the sizes of each candidate's "hump" and, in a two-person race, it's pretty trivial to use this to estimate probabilities. There's virtually no quantitative data in the universe outside of possibly Sanders' campaign office that could predict this outcome. Don't shit on people for not predicting it after the fact. The performances of the campaigns of both sanders and trump suggest that folks like silver need revisit the degree to which their models provide for whatever falls outside the realm of quantitative data, which sort of flies in the face of what 538 is all about. This ain't 2012;) Edit: and yep, on spring break this week, though I vote in Ohio 😊
I don't think that's really true. They could integrate closed vs open primary a little more and perhaps build in more uncertainty, but I highly doubt anyone could craft a model with better performance than theirs thus far in the primaries given the paucity of data (and if they have, kudos to them).
You could build a model with so much uncertainty that it can handle something literally unprecedented in the history of modern polling like Michigan today, but then it wouldn't be very useful (which is kind of the point of a model since they're never "right" anyway).
|
United States22883 Posts
Also, it does not surprise me that Kasich is getting crushed here. Republicans in Idaho are all sorts of fucking crazy.
Case in point: This piece of shit who shot a pastor, from Cruz's rally, in the head, and then drove across the country and attempted to assault the White House.
http://www.cnn.com/2016/03/08/us/idaho-pastor-shooting/index.html
Keep in mind a different Idahoan was put in prison for firing at the White House in 2011.
I always say the worst part of Idaho are the Idahoans.
|
On March 09 2016 13:33 kwizach wrote:
The A.P. calls Michigan for Sanders! Kudos to his campaign and their volunteers for grabbing that one and making the polls lie.
Thanks we busted ass for it, and it wasn't easy getting motivated with practically every Hillary supporter and Media outlet saying we should of given up.
But this is why we didn't, it wasn't over. Nice try though.
|
Haha so media outlets on why they were so wrong? It's because Michigan has had such odd primary's.
Also if Clinton can't win the Rust Belt that means she has to pick either Warren or Brown as a VP pick. Smart pick would be Brown an have an ally in Warren in the Senate.
|
On March 09 2016 13:33 kwizach wrote:
The A.P. calls Michigan for Sanders! Kudos to his campaign and their volunteers for grabbing that one and making the polls lie.
Just want to say thanks for being gracious in defeat.
yooj night
|
On March 09 2016 13:37 Jibba wrote:Show nested quote +On March 09 2016 13:33 ZeaL. wrote:On March 09 2016 13:28 TheTenthDoc wrote:On March 09 2016 13:23 ZeaL. wrote:On March 09 2016 13:20 Jibba wrote:On March 09 2016 13:16 ZeaL. wrote:On March 09 2016 13:12 GreenHorizons wrote:On March 09 2016 13:05 ZeaL. wrote:Why is Gladwin County 780-74 for Sanders with 17/18 precincts reporting? Edit: Tweet from 538 earlier today: + Show Spoiler + 538 totally air-balled MI, no if ands or buts about it. They didn't even skim the net let alone touch the rim. I mean they were totally off but it's not completely their fault. Polling for MI has consistently had Sanders down by 10-20% within the last week even. They can't do anything without accurate polls. Even if Hillary pulls it back and somehow wins, this result still shows the polling was way off. They can adjust based on the quality of polling data. Silver did it before, and it's why he ran counter to mainstream polls. They can, but if every single poll is saying Clinton by 10-20% then no matter how you weight the data you're not going to get an expected Sanders win. Edit: I guess if they know the polling data is poor they should have broad confidence intervals. For some reason they neglect to include them in their model. Their model already has confidence intervals. The uncertainty around the point estimate is reflected in the sizes of each candidate's "hump" and, in a two-person race, it's pretty trivial to use this to estimate probabilities. There's virtually no quantitative data in the universe outside of possibly Sanders' campaign office that could predict this outcome. People shouldn't shit on people for not predicting it after the fact. Yeah thinking about it a bit more I understand why the output is presented the way it is. And just for reference this is the data they were working with: If anything this is an indication of the poor quality of polling companies. They've always been poor quality polling. We all understand garbage in, garbage out. Nate's magic last time was the system for rating polling quality. The actual vote model probably wasn't that much different than MIT's model.
Correct me if I'm wrong but I recall from the '08/'12 elections the idea was to add a bias to the poll (i.e. Gallup is +R so correct by -R +D) and then weight the value of the poll based on its quality (trust historically accurate polls over unproven/inaccurate). This seems okay for general election polling but how would you apply this to a primary given no historical data between the candidates?
|
|
|
United States22883 Posts
On March 09 2016 13:44 ZeaL. wrote:Show nested quote +On March 09 2016 13:37 Jibba wrote:On March 09 2016 13:33 ZeaL. wrote:On March 09 2016 13:28 TheTenthDoc wrote:On March 09 2016 13:23 ZeaL. wrote:On March 09 2016 13:20 Jibba wrote:On March 09 2016 13:16 ZeaL. wrote:On March 09 2016 13:12 GreenHorizons wrote:On March 09 2016 13:05 ZeaL. wrote:Why is Gladwin County 780-74 for Sanders with 17/18 precincts reporting? Edit: Tweet from 538 earlier today: + Show Spoiler + 538 totally air-balled MI, no if ands or buts about it. They didn't even skim the net let alone touch the rim. I mean they were totally off but it's not completely their fault. Polling for MI has consistently had Sanders down by 10-20% within the last week even. They can't do anything without accurate polls. Even if Hillary pulls it back and somehow wins, this result still shows the polling was way off. They can adjust based on the quality of polling data. Silver did it before, and it's why he ran counter to mainstream polls. They can, but if every single poll is saying Clinton by 10-20% then no matter how you weight the data you're not going to get an expected Sanders win. Edit: I guess if they know the polling data is poor they should have broad confidence intervals. For some reason they neglect to include them in their model. Their model already has confidence intervals. The uncertainty around the point estimate is reflected in the sizes of each candidate's "hump" and, in a two-person race, it's pretty trivial to use this to estimate probabilities. There's virtually no quantitative data in the universe outside of possibly Sanders' campaign office that could predict this outcome. People shouldn't shit on people for not predicting it after the fact. Yeah thinking about it a bit more I understand why the output is presented the way it is. And just for reference this is the data they were working with: If anything this is an indication of the poor quality of polling companies. They've always been poor quality polling. We all understand garbage in, garbage out. Nate's magic last time was the system for rating polling quality. The actual vote model probably wasn't that much different than MIT's model. Correct me if I'm wrong but I recall from the '08/'12 elections the idea was to add a bias to the poll (i.e. Gallup is +R so correct by -R +D) and then weight the value of the poll based on its quality (trust historically accurate polls over unproven/inaccurate). This seems okay for general election polling but how would you apply this to a primary given no historical data between the candidates?
The weighting is applied to the polling service, not the candidate though.
|
It seems to me like Trump is approaching being the inevitable candidate. Is there really a way for anyone else? Seems like he pretty much has it locked up barring a meltdown.
|
On March 09 2016 13:50 Slaughter wrote:
It seems to me like Trump is approaching being the inevitable candidate. Is there really a way for anyone else? Seems like he pretty much has it locked up barring a meltdown.
Rubio is pretty much screwed now. It'd take a miracle but I'd say there's still a small small chance the party decides to back Kasich and Cruz falls apart.
|
GOP side is basically between Cruz and Trump. Establishment needs to bite the bullet and pick one. Even if Kasich wins Ohio or Marco wins FL, it doesn't matter much. Both are looking less and less likely
|
On March 09 2016 13:48 Jibba wrote:Show nested quote +On March 09 2016 13:44 ZeaL. wrote:On March 09 2016 13:37 Jibba wrote:On March 09 2016 13:33 ZeaL. wrote:On March 09 2016 13:28 TheTenthDoc wrote:On March 09 2016 13:23 ZeaL. wrote:On March 09 2016 13:20 Jibba wrote:On March 09 2016 13:16 ZeaL. wrote:On March 09 2016 13:12 GreenHorizons wrote:On March 09 2016 13:05 ZeaL. wrote:Why is Gladwin County 780-74 for Sanders with 17/18 precincts reporting? Edit: Tweet from 538 earlier today: + Show Spoiler + 538 totally air-balled MI, no if ands or buts about it. They didn't even skim the net let alone touch the rim. I mean they were totally off but it's not completely their fault. Polling for MI has consistently had Sanders down by 10-20% within the last week even. They can't do anything without accurate polls. Even if Hillary pulls it back and somehow wins, this result still shows the polling was way off. They can adjust based on the quality of polling data. Silver did it before, and it's why he ran counter to mainstream polls. They can, but if every single poll is saying Clinton by 10-20% then no matter how you weight the data you're not going to get an expected Sanders win. Edit: I guess if they know the polling data is poor they should have broad confidence intervals. For some reason they neglect to include them in their model. Their model already has confidence intervals. The uncertainty around the point estimate is reflected in the sizes of each candidate's "hump" and, in a two-person race, it's pretty trivial to use this to estimate probabilities. There's virtually no quantitative data in the universe outside of possibly Sanders' campaign office that could predict this outcome. People shouldn't shit on people for not predicting it after the fact. Yeah thinking about it a bit more I understand why the output is presented the way it is. And just for reference this is the data they were working with: If anything this is an indication of the poor quality of polling companies. They've always been poor quality polling. We all understand garbage in, garbage out. Nate's magic last time was the system for rating polling quality. The actual vote model probably wasn't that much different than MIT's model. Correct me if I'm wrong but I recall from the '08/'12 elections the idea was to add a bias to the poll (i.e. Gallup is +R so correct by -R +D) and then weight the value of the poll based on its quality (trust historically accurate polls over unproven/inaccurate). This seems okay for general election polling but how would you apply this to a primary given no historical data between the candidates? The weighting is applied to the polling service, not the candidate though.
Right and that's fine if the polling services disagree, heavily weight the accurate ones and figure which result is more likely. For IA, NH, NV, SC there were all at least some polls close to the final result so modeling based only on polls could predict the actual winner. For MI though there were literally no polls showing Sanders even close to a win. There's no way you can twist the data to get a Sanders win out of that unless you assume every polling outfit is heavily Clinton biased and add a big fudge factor to your data.
It's just weird that to me that polls could be relatively accurate for the first few states and then be completely off. There has to be something that is not being accounted for. If 538 wants to continue to be the oracle of elections then I hope they figure out what is really predictive.
|
After SC, there were many Cruz supporters who were thinking of backing Rubio under the (ultimately false) establishment narrative that Rubio had the best chance. After Super Tuesday, no one was thinking that anymore- it was quite obviously wrong. Whatever chance Cruz has, Rubio's is clearly worse.
The reason Trump is so likely is that either A) he wins Florida/Ohio, and simply has too many delegates or
B) Kasich/Rubio win their respective states, so no one drops out, and Trump continues to dominate a crowded field.
This craziness is only helped because the upcoming states are better for Rubio (and maybe Kasich), but it's already too late for him/them.
So do they go all out and play for a contested convention? That's what they are deciding now, if they weren't discussing it already. After all, rumors have been going around for at least a week now.
|
As a Buckeyes fan, I have to say...Go Michigan! I had a feeling he would do well there. Once you get out of the South, I really think you'll start to see that there is plenty of minority appeal for Bernie. The narrative has been so jaded since so many of the early voting states have been concentrated in that area. Hopefully the rest of the country can continue the momentum 
|
General overview of polling methodology in depth: http://www.pewresearch.org/methodology/u-s-survey-research/
I read somewhere, not sure where, that pollsters have been aware for a long time that their sampling methods are faulty for producing random samples. The only reason polls worked is that for a reason they were unable to explain, telephone poll results seemed to track relatively well to actual results.
Granted, that has pretty much fallen apart this election, and Michigan here is one of the most notable examples.
|
Well this continues Rubio's terribly awful streak. Now I can see why Cruz was floating the idea of moving into Florida-no way Rubio can beat Trump there after tonight short of killing him.
|
|
|
|
|
|
EPT
