Okay, I haven't seen anyone yet do what, to me, would be the most conclusive evidence for or against voter fraud, and a statistician with the right software could probably do the trick without too much effort.
I think it would be easy to compare Bush's tallys in states that used software-based touchscreens and those that used more traditional, tamperproof methods (punchcards, optical scans, etc.)
Then one would just need to compare the results across a couple of variables (party ID, 2000 support, turnout percentages, etc., Congressional elections, etc.) and I think you'd have an answer to the vote fraud question.
Any takers?