I saw this in the bill of complaint filed by the state of Texas (?!) challenging results of elections in PA, GA, WI, and MI. I knew I had to know what sophisticated statistical methods were being put to use. So CHUCKLE IN BUCKLEHEADS it's time to talk about DIFFERENCE OF MEANS

The "expert" testimony is by Charles Cicchetti, an economist who appears to specialize in energy and environmental issues. Why he feels comfortable commenting on elections is unclear, as he has clearly never read anything about elections in either econ or poli sci

Cicchetti is responsible for the claim that the chances of Biden winning four states that Trump won in 2016 are about one quadrillion to the fourth. Where does this claim come from? He makes two main arguments.

For one, he compares the vote totals of biden compared to the vote totals of clinton in each state. He does a difference of means test. What this means is he's subtracting the number of votes Clinton got from the number of votes Biden got and dividing by a standard error.

That number is what he calls a "Z-statistic." He then uses a computer to tell him what percentile this is on a z distribution and comes up with p=basically 0 but calls it p=0.00000000000000000000000000001 to make a point (I did not count these 0s).

So what does this statistic tell us? Well he says it: "I can reject the hypothesis that the percentage of votes that Clinton and Biden achieved in the respective elections are similar."

I can be a little more specific: the difference of means test is a test of population parameters in two samples. If you take two random samples from one population, both samples should, if they are large enough, reflect the population parameter.

So if you took 2 random samples of 100 american adults, and run a dif. of means test on average height in each sample, you should find they have no significant difference. But if they were from 2 populations with, e.g. different levels of nutrition, you may find a difference

So what the esteemed Dr. Cicchetti found is that, 1. if you treat who voted as a random sample then 2. the populations of voters in 2016 and 2020 are not the same

Which 1. lol it's not and 2. lol we know

People don't vote at random! And there are changes between elections! That's why the same party doesn't win every time!

This guy literally did a bunch of tests to find out that the margins in 2020 were significantly different than the margins in 2016.

So his one in quadrillion whatever chance is not the chance of biden winning or whatever. He does nothing to even approach that as a question. It's the chance that the number of people in the population that support biden is the same number that supported clinton

(If people vote at random. Which they don't).

Anyway he has more stuff about the time of reporting that is along similar lines but I'm tired. May take it up later.

I have to keep the motif of the thread though. YOU JUST LEARNED SOMETHING YOU MORONS.
n/n