General Discussion
In reply to the discussion: Wisconsin: None dare call it vote rigging [View all]Igel
(35,300 posts)We cry foul when the final numbers are out of line--and almost always where there's independent observation of what went on.
The problem is that the pollsters do exactly what the guy who took a stat class for PhD students in the Social Sciences say they don't do. He assumes they just "adjust the data" to match the final polls.
They adjust the data in a number of ways, and do it precinct by precinct. That's not matching it up to a single data point, that's matching it up to hundreds and hundreds of data points. You sample certain precincts, you know how they voted. That lets you deal with all the non-respondents. In the past, it's pointed out problems with the survey protocols, it's pointed out sloppy recording, it's currently pointing out that the pollsters had a bad model for interpreting the early data.
Early voting and absentee ballots aside, it's what you need to do. Yesterday there was a lengthy post concerning a racial bias in Gallup's polling methods pointing out exactly how you have to weight (opinion) poll numbers and all the reasons for it, that you can't use "unadjusted numbers". People like the idea that Gallup was misweighting numbers, and all over the idea--in support--that you have to adjust numbers. Mostly, it seems, because they liked the direction the adjustments went in. Not because they believed the math required it, but because it favored them. (Critical thinking? Just say no!)