Never seen this
Posted on 09/26/2016 6:21:25 AM PDT by Ravi
People over at The_Donald have pretty good info about Nate changing the methodology by changing his “house effect” to keep Hillary ahead.
https://www.reddit.com/r/The_Donald/comments/54kkhs/pathetic_nate_plastic_from_538_btfo_caught/
You are right in saying that anybody can do 50-50, but that isn’t all Nate does. His strength isn’t so much in the 50-50 realm but in the 67-33 to the 95-5 realm. A 67-33 prediction should be correct only two-thirds of the time, not more, not less. The outsomes are not independent of eachother, which gives another level of complexity.
Hillary’s odds of pulling a state that Mittens won have dropped to 50%, which means that some states—mostly AZ, NC, and GA—while on average coming up red, come up blue often enough that half of his simulations have one of them blue. Her odds of winning are only 1.5% above the flipping a Mittens state level, which means that in most of the simulations where she didn’t over-perform in one of these states, she lost.
An August poll had Clinton leading 40%-31%, but an Ipsos poll in early September had Trump leading 43%-38%. New Mexico needs a fresh poll.
-PJ
Fact is by the end of the week even the Liberal media will have to realize with a little over a month to go, Trump is riding a wave to victory. Hillary has thrown in the towel on Ohio, will probably give up on NC and Florida the next week to ten days and spend most of her time in her Stalingrad....Va, Pa, Mi and Wi. I think a real shocker on election night, to the horror of the MSM, is that most of the upper Midwest will go Trump. I suspect the trifecta of Wi, Mi, and Ohio. Trump will also lose NY, NJ, Ill. and Ca by less than a couple of points. I'm thinking he wins with right around 300EVs.
I am still a little negative on Pa, tons of voter fraud and too many white trash idiots and minority chattle in the Burgh and Philly. I do see Hillary signs here in Pitt, in the black communities and white Appalachia bordering places like Uniontown where the wigger boys wear pants around their knees and you see lots of 350lb white welfare queens with their Lil Wayne BFs.
Exactly, he wants an accurate outcome. He’s not trying to affect the outcome by playing with his data.
He recooked the state polls today. The proof is in my #21
Libs are giving the guy a lot of grief on Twitter
Co and NM showing Trump, which wold flip Nate’s map.
HRC over 50% though? That’s scary to me
-PJ
I love it when a plan comes to together.
MSM is panicking and the Washington Post’s wheeling out an editorial blasting Trump ahead of tonight’s debate tells you the election is over.
I’ve never seen liberals as panicked as they are now. What a beautiful sight! Things are looking up.
Almost 70% certainty, in Silver’s world, isn’t very certain, and while he may like Democratic demagogues and free wine from liberals, I think he likes being right more.
By the way—I think that if trump looses Ny, IL and CA by only a couple of points, he wins NJ. I’ve been keeping an eye on where NJ is in Silver’s esitmation in comparison to Trump states. Right now he has Trump taking NJ slightly more often than Clinton takes Georgia. RI and DE are even more likely.
Worse case for Clinton is that the early results are so crushing that wonky things happen to the west coast turn out.
Map is out of date.
Trump has taken the lead in CO.
You can bet he’s up in NM as well.
Trend is your friend.
Nate Silver’s model is nothing more than a weighted average of all available poll data. The data is adjusted in two ways, but the adjustments are intended to promote increased accuracy, not favor either outcome. One adjustment is for house effect. There are some polls that historically have shown results more favorable to Republican candidates than the actual vote totals and some that are historically more favorable to Democrats. Silver computes an average value of this effect for each poll and adjusts the poll number accordingly.
The other adjustment is for the quality and track record of the poll. Even after adjusting for house effect, some polls have been more accurate than others historically. Since the average is a weighted one, Silver assigns weights to each poll. Those historically more accurate are assigned higher weights. Historically less accurate polls, new polls with no track record, and polls that use questionable methodologies are assigned lower weights. Polls done more recently also get higher weights than older ones.
The end result is that Silver’s opinions don’t factor in at all during the course of the campaign. Certainly the weights assigned to polls are subjective and do reflect Silver’s opinion regarding the quality of the poll, but these weights are set based on a polling company’s track record and methodology. This is done prior to the campaign, and the weights are set indepemdently of the poll result.
We may not have liked Silver’s predictions the past two Presidential campaigns, but the truth is that his predictions were just about dead on. Blame Obama and a couple of weak GOP-e candidates for that - it’s certainly not Nate Silver’s fault we lost. Given his record the past two Presidential elections, I am certainly taking his forecasts seriously. There’s no guarantee he’s right, but there’s also no evidence that he’s biased either.
Never seen this
But all of this assumes that the underlying data is correct, even if the electorate modeling is not applied properly.
In this election, I think there is a certain number of respondents who are not revealing their true choice. No amount of correction of the data can account for that. And that's one reason why I think Trump's numbers are understated.
There are more complex ways of adjusting polls than always adding or subtracting in one direction. Silver speaks of the problems of herding. He may regress outliers to the mean somewhat. He hasn’t made his formula public.
I don’t think he merely looks at a poll and says “let’s move it a couple of points this way or that.”
At any rate, if he is trying to stack the deck, stackig it so that Hillary still looks to be in bad shape doesn’t seem effective. My bet is that Silver wants to be right, and however he gets there, he is going to get as close as possible to what he thinks will be the correct outcome by election day.
He had an article early this morning noting that the final pre-debate polls were generally more predictive of the final polling before election day than the post-debate polls.
-PJ
PING
Whoever carries FL and OH wins the election.
Pray America wakes
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.