Comment by deckar01
5 months ago
I was able to reproduce the vote skew / batch size correlation in early voting. The average batch size for mail voting was way higher (6 machines) and election day was lower (3x machines). The fact that the vote ratio is more distributed for mail voting with way more votes per machine is what highlights incongruity for me. Stats are not always intuitive though.
Edit: Mail voting is not more distributed. You can’t analyze the distribution of a set that small using its min and max, it has to be the full range like the other sets. Distribution approaches zero as bin size increases regardless of voting source. This no longer seems surprising.
Yeah, I was confused about that when I tried to replicate the study for all of the counting groups ('Mail', 'Early Voting', 'Election Day'). I'd like to know more about the data to understand how to compare the others with early voting.