Her analysis is startling. Certain counties voted for Bush far in excess of what one would expect based on the share of Republican registrations in that county. They key phrase is "certain counties"--there is extraordinary variance between individual counties. Most counties fall more or less in line with what one would expect based on the share of Republican registrations, but some differ wildly.
How to explain this incredible variance? Dopp found one over-riding factor: whether the county used electronic touch-screen voting, or paper ballots which were optically scanned into a computer. All of those with touch-screen voting had results relatively in line with her expected results, while all of those with extreme variance were in counties with optical scanning...
Check it out and see if you think his analysis holds up or not.