Yeah, I’ve seen those and a few others.
OTMAL had a good point, that the leading digit distribution can vary based on size of precinct. In other words, that Benford’s law has limitations. Applications where the results it provides do not justify confidence in a conclusion. In particular, a conclusion that fraud is present.
OTMAL pointed out some facts - Miami curves do not diverge, other cities they do diverge. Same clinical observation you provide.
I was seeking OTMAL conclusion. Given that Benford “breaks down in some cases,” does it break down in THIS case. Is there evidence of fraud, or not? I know my gut feel, I know your position, but OTMAL pointed out a weakness on Benford. I wonder what effect that has on his confidence in the conclusion that fraud is present.
I’m fine with statistical methods in general, all sorts of distributions, assumptions, etc. Generally though dealing with physical qualities that have variation. Benford is not part of my study experience, and I was looking for an explanation of ramifications of what looks to be a “be careful what you conclude” admonition from OTMAL.
I saw another layout of graphs from MI that were based on precincts comparisons not leading number theory.
As examples of vote fraud they were irrefutable. Bidens numbers were a broad sawtooth not a curve while all trumps numbers were within 5% of mean.
These anomalies showed up in the city centers that were barring access to poll watchers, and the site that had a witnessed late night ballot dump with unofficial vehicles after hours.
Note that software block shifting of small pecentages would tend to be not detected by this method as the shifts would be percentages of natural numbers, not human entry.
I suspect the vast majority of voting machines are set to subtly bias and create “the blue wave”, and this amateur ballot drop was because the bias was set to low due to overwhelmingly massive Trump turnout.
Commandering a few ballot machines and doing a forensic software study is in order.