Quote (bogie160 @ Sep 17 2016 05:58am)
He wrote a piece specifically about it.
He thinks it has a Trump house-effect, which it almost certainly does based on the way they decided to weight respondents, but he likes that it follows the same group from poll to poll and he thinks it's a useful barometer, once the house effect has been taken into account, for how the race is progressing.
The race has clearly tightened, there are several new avenues for a Trump victory that didn't exist a month or so back.
I guess you didn't read that piece very clearly -- neither it/he/any of what you wrote disagreed with what I wrote originally -- you just restated it. Exactly as I said, if he thinks he can glean value from it by including it in his model then that's his choice. It's still a panel. Panels can be incredibly useful when executed properly (this one however is dogshit). It obviously has incredible flaws as I mentioned; if he feels like he can remedy them through adjustment then he's free to try that, he's built a model that calls for including basically any and all available polls with the belief that his model can "fix" the bad ones. Regardless of how the model performs, it's still garbage in/garbage out; he can't pretend the survey doesn't have immense flaws or that it's not a panel. He's simply saying implicitly by including it "no big deal, these are trivial concerns that can be corrected."
I think people really need to understand this point: the race hasn't tightened,
some public polls have tightened. Trump might honestly have exactly
zero additional avenues for victory that he didn't have one month ago: he was always going to need to win Iowa, Ohio, Florida, keep North Carolina from getting poached, and then push into the ring of WI/CO/VA/PA/MI that he has had nothing going for him so far. That was true then and it's true now.