The overwhelmingly interesting thing I noticed here was that everyone seemed to accept—not explicitly, but implicitly very much—that an Obama supporter acting violently was in some sense evidence against Obama or justification for opposition to Obama; or, that a McCain supporter acting dishonestly was in some sense evidence against McCain or confirmation that Obama supporters were better people. To a Bayesian, this would be balderdash.
It would be Bayesian evidence of the right sign. But its magnitude would be vanishingly tiny.
It would be Bayesian evidence of the right sign. But its magnitude would be vanishingly tiny.
Considering how many ways either outcome would result, im not really sure how P(supporter carves a B |obama is evil) would actually measure out