I’m know you’re complaining on Twitter about people who aren’t reading your piece, but it seems you didn’t read 538’s piece very closely and if you did it’s clear you don’t understand it. For your link on evaluating the whether pollsters are consistently missing in favor of one candidate or the other:
“In the Democratic race, the polls had a 1.8 percentage point bias toward Clinton (and against Sanders) overall. However, it varied significantly based on the demographic makeup of the state, with Clinton outperforming her polls in diverse states and Sanders beating his in whiter ones. Specifically, in states where at least 25 percent of the population is black or Hispanic,4 the polls had a pro-Sanders, anti-Clinton bias of 5.7 percentage points. But they had an 8.2 percentage point bias toward Clinton, and against Sanders, in states where less than 25 percent of the population is black or Hispanic.”
A 1.8% of bias toward Clinton is far less than the 10.6% you’re throwing around and it’s nowhere near high enough to assume Bernie still has a chance of winning if underlying factors don’t change. (You do consider this possibility, but ignore the chance of Bernie making his own gaffe that could hurt him, Trump finally attacking him in full force, or that Clinton could do something very good that draws support toward her. This applies to both voters and superdelegates.)
But the more egregious issue with this screed is that it badly mangles the idea of “margin of error” and completely misapplies it. First off, the 10.6% you cited was never the margin of error, it was the average error of individual polls. Margin of error is a specific calculation in which the error is a fixed value based on sample size. Furthermore, if you’re sampling a very small group (such as ~700 superdelegates) you have adjust your standard error down to account for the larger sample percentage (the finite population correction). So you can’t just take the standard error (which would be the correct “error” to calculate here) from a sample from millions of people and apply it to sample from about 700 people.
Of course, all of this is moot because I can find the preference of *every* superdelegate online. I don’t need to sample at all (which is why there are no superdelegate polls akin to voter polls) making it *impossible* to have a sampling error..
And no, you can’t use specific statistical terms however you want because they have very specific meanings. The fact a superdelegate may change his or her choice does not have anything to do with errors in polling. I saw you trying to make the argument on Twitter and it just obscures the breathtaking misunderstanding you have of statistics, which you have been repeatedly and correctly called out on.