While EDA still thought they had an "uncanny reversal" statewide, they looked at the vote percentages by county, which are all over the place.

Therefore, the hypothesis that vote reversal took place in all machines goes away. Incidentally, that would have provided fraudsters with plausible deniability: "oops, programming error".

However, looking at the percentages and the sizes of the counties, we see that the largest county, Hillsborough, shows a very large near reversal of the percentages. Therefore, EDA has suggested that fraud might have tafen place only in Hillsborough.

Comments on that. The behaviour of the 10 counties is all over the place. Grafton went for Obama more strongly on machine-counted precincts than it dit on hand-counted precincts. Clinton won Coos more strongly on machine than on hand-counted precincts. In Stafford, both of them gote a higher vote share on machines than they did on hand counts.and so on. Now, given that the behaviour of the 10 counties is all over the place, the requirements for statistical significance for having a single county with a reversal are multiplied by 10. That is, the reversal needs to be very much closer to exact than otherwise. Especially because Hillsborough has been identified from looking at the the data. When you test a hypothesis formulated after looking at the data, the requirements for accepting it are stronger. I suppose the next step here is to do the same plot by municipality/ward within Hillsborough county.

We have met the enemy, and he is us — Pogo