Read on Twitter
I think the correct answer to Paxton's expert's claim is as follows. Most frequentist statistics is dedicated to answering a very specific question: What are the odds of something occurring *by random chance*? 1/
That, after all, is effectively what a p-value is trying to tell you: If there were no relationship between variable A and variable B, cet. par., what is the probability that we would nevertheless see the sort of data that we are observing. 2/
So, yes, it's extremely unlikely that Trump's substantial lead on election night would disappear by random chance. It may even be a quadrillion-to-one. But statistics can go no further than that. Now we must ask "with random chance ruled out, what are our options?" 3/
As it turns out, in this particular instance, excluding random chance does nothing for you, because everyone (well, almost everyone) agrees that ballots are cast or counted at random. 4/
We had one presidential campaign very much dedicated to ensuring that their voters would cast ballots by absentee voting, and another very much dedicated to the opposite. 5/
IOW "it's extremely unlikely that this would occur due to random chance" doesn't lead to the conclusion "it was therefore fraud" at all. There's an extremely good, non-conspiratorial explanation that doesn't depend on chance, and there's every reason to believe it. 6/6
