General Discussion
In reply to the discussion: WSU statistician sues seeking Kansas voting machine paper tapes [View all]HereSince1628
(36,063 posts)It's actually conceptually very simple...
when sample sizes are low, variance should be large and noisy, but as sample size increases, the variance will settle down...exponentially approaching a constant as sample size increases.
It's a basic feature of statistics long used to determine minimum sample size...
So if you look at voting returns as they are counted...the variance at first should be large and noisy as the count goes towards completion the variance should settle to a constant.
That early study found that variance did NOT settle down as would be expected for a random process, rather it sometimes increased and in many cases decayed linearly rather than exponentially.
And it happened on in large enough precincts to cause small percentage changes that turned precincts red. The odds of it being manipulated were over a million to one.
And Walkers wins in WI are all within 3% swings.