Jump to content

DMC

Members
  • Content Count

    13,088
  • Joined

  • Last visited

1 Follower

About DMC

Profile Information

  • Gender
    Array

Recent Profile Visitors

12,167 profile views
  1. DMC

    US Poll-itics

    I know this is a whole day old now, and isn't going to change anything, but I don't think I've seen it mentioned on this thread and it's preposterous enough to share: Trump telling women voters he's going to get their husbands back to work:
  2. DMC

    US Poll-itics

    A 30 percent chance - which is what the final 538 forecast gave Trump - is a pretty damn good chance. As for "all the models being off in the same way," well yes, all the polling the models were based off of would have to be off in the same way to elicit a significant polling error - otherwise they'll cancel each other out and there won't be an aggregate polling error. In 2016 it went Trump's way. In 2012 it went Obama's. I haven't said Biden would comfortably win. It's like you're incapable of understanding nuance. I've said if Biden maintains the current lead in aggregate polls - both nationally and in swing states - it would require a massive systemic polling error that would be significantly larger than the 2016 error. I also have said I agree if Trump loses closely, which would inherently only require a more muted polling error, I agree he'll be successful in stealing the election with the help of the SC. I've also mentioned - albeit I don't think to you directly - that I don't take 538's current probabilities of a Biden victory at face value. I do not think Biden actually has an 88% chance of prevailing - and I've said before I doubt Silver actually does either - because of all the factors you highlight and I tend to agree with, just not at the magnitude you do.
  3. DMC

    US Poll-itics

    Probabilistic models are not predictions. Silver has been trying to emphasize this for the past four years. As for you being "in line" with Silver, that's especially rich even for you.
  4. DMC

    US Poll-itics

    Which demonstrates either your ignorance or intentional bias - there were plenty of (virtually all) other modelers and pollsters showing a much greater chance for Trump in 2016 than the Princeton model.
  5. DMC

    US Poll-itics

    I'd argue wholly dismissing the metrics and analysis of people that devote their lives trying to figure out what's going on and instead relying on your gut feelings is "uncurious," by YMMV.
  6. DMC

    US Poll-itics

    Well, in that case, I've said multiple times that if Biden maintains his current lead it would require a massive systemic polling error which I think is very unlikely. I've also said multiple times that as long as Biden wins convincingly, Trump will be unsuccessful in his attempts to change the election results via court challenges and GOP officeholders will abandon his efforts. I've also said multiple times that if the election is close enough - e.g. like Hillary in Trump losing the EC by 100-200k votes - I agree with you it's very likely Trump will be successful and the SC will find a way to hand him the election. I have said the latter two things multiple times specifically in multiple discussions with you over the past few months. If that makes me "cowardly" in your book, well then I'm proud of that.
  7. DMC

    US Poll-itics

    No, when analyzing human behavior the environment is inherently going to be different in each "experiment" outside of the lab. And like I said, all elections are unique and you could make a compelling argument that virtually every election involves such different conditions to render all prior data moot. In 2016 it was the first time a woman was the major party nominee, the other nominee was the first with no prior political/military experience, and the two of them were the least liked major nominees in the history of polling. In 2008, we had the first major party black nominee and the greatest economic downturn since the Great Depression (which preceded polling). 2004 was the first post-9/11 contest and we were in the midst of the most significant war since Vietnam. Like I said, shitting on all of social science is your prerogative, but it is also decidedly anti-science. As for "controlling the variables," no, I've never heard that phrase. What I do know of is including control variables that are fixed, which all the models you're referring to do employ. Except it's only "bad data" in your warped perspective, and in actuality is based on well-founded data that, generally, has elicited probabilities that correspond to the outcome with marginal errors. Ha! So, one side of your mouth is spewing horseshit about how these data and subsequent models are entirely useless, and on the other side you're calling me cowardly for "not sticking my neck out" based on such models? What exactly do you want me to stick my neck out in saying?
  8. DMC

    US Poll-itics

    The "degree of error" is statistically determined by the model and the volatility of the variables. Obviously, other than lab experiments, all social science cannot recreate the controlled environment of the hard sciences. If you want to do dismiss all non-experimental social science because of this, that's your call. Variables, by definition, don't stay the same. Never said that... Never said I was confident in any result either. You, however, have suggested multiple times in the past 24 hours that essentially all polling is pointless because of all the differences in this election - and explicitly stated these differences render all prior data useless. That is not remotely the same as "talking about a +2 movement" and is inarguably an anti-science perspective. Enjoy your pathetic trolling.
  9. DMC

    US Poll-itics

    Here's an article detailing QAnon's obsession with adrenochrome. By far, my favorite part of the article:
  10. DMC

    US Poll-itics

    Easiest way to do this would be to use automated textual analysis for a selected amount of blogs, tweets, etc. Establish a standard for using text analysis techniques (e.g. frequency, classification, concordance, clustering of keywords), then integrate this with structured data (i.e. polls, demos, datasets - the variables that compose all current models). This type of design is more and more frequently being employed by researchers throughout the discipline (my advisor made me do it for one of my diss papers). However, I'm quite skeptical how much added value you'd get from the text analysis. A lot of the blogs and tweets are just going to be feedback so all you're really capturing is the echo-chamber of (elections) data journalists and their political junkie followers. Ha, no. Is this election unique? You bet. So is every other election. That doesn't make constructing models based on prior election data "useless." Not to mention you're just basically wrong in most of the last sentence - this is hardly the first election with early voting, is not even the first election since Shelby, and definitely not the first election where at least one candidate is trying to employ voter suppression to steal the election. This entire perspective is anti-science; "we can't construct models because this election is so different than previous elections." Yeah, that's why they're called variables. Is this election more different than the differences between other elections? Sure! And the volatility of, say, the meteoric rise in early/mail-in voting would be reflected in the statistical uncertainty of that variable in the model. But, I know, all polls and all of us trying to make sense of what's happening in this election based on data are just laughable peons engaged in a futile endeavor, relying on well-founded metrics that pale in comparison to the sage wisdom of the The Kal Gut - which knows better than all. Perhaps instead of constantly trolling those of us trying to analyze the real world, you'd be more comfortable with your own kind - like Trumpists decrying that all polls are wrong, the climate change deniers, the covid hoaxers, etc.
  11. This is what Orlando is for.
  12. DMC

    US Poll-itics

    Heh, can't argue with that logic.
  13. DMC

    US Poll-itics

    Proper methodology still counts!
  14. DMC

    US Poll-itics

    Just curious, is there any reason you've shifted back to cynicism other than Barrett being confirmed? Definitely not giving up hope for Florida. Hell, in 2012 - based on Florida's RCP final average I posted upthread - the polls were off in Obama's favor by 2.4%. I'm just pretty cynical about the state after 2016 and 2018 - and the fact they got a bunch of Trumpist hacks at the top levels of state government/overseeing the election. It also personally pisses me off since my parents, sister and brother in law live in Gainesville and Tampa, plus I lived in Orlando for 7 years (and I still vote there). Yep, looking at the numbers of ballots returned versus requested is entirely the wrong metric to use. In 2016 only 69% of ballots requested were returned nationwide. Who knows if that number will go up or down, but gun to my head if I had to guess my prior would be it should go down simply based on how many more people are able to easily request a mail-in ballot. Now this would be a much better metric to use - albeit still flawed since it was a primary AND this was at the beginning of covid when everything was quite hectic for voters and states (whereas each have had much more time to prepare for the general). However, the good thing about Wisconsin's April 7 election is it wasn't just a primary - there were also judicial and local elections. Not sure where you're getting that 80k number, but if it's based off of the total amount of votes cast - 1,551,711 - that's about 5 percent of the total vote. So let's take that number and assume the cutoff ballots would go 2 to 1 for Biden. That would mean Biden would lose out on a 1.7% margin. So, if Biden's up by 5 in PA, that would mean he only has breathing room for about a 3 point polling error. That totally sucks, but still a very good chance that can be overcome - and I think those (rather baseless and wild) assumptions are on the highest end of how much such a cutoff would hurt. Obviously, Barrett's confirmation likely meant Dems' approval of SCOTUS took at least decent hit. Be nice if there coulda been polling on that, but they're kinda busy right now. However, I don't see any reason why that hit should be much more pronounced than the hit it took when McConnell refused to take up Garland's nomination. Plus, gotta figure there'll be a boost from GOP respondents that are happy with the Barrett confirmation. Regardless, I'm not interested in the future - or at least anything beyond deciding this election. And really the only thing that matters in terms of the court's legitimacy in the context of this conversation is that Roberts currently thinks the court holds legitimacy and wants to maintain that. I'm very (very) confident both of those are the case.
×
×
  • Create New...