csc@watmath.UUCP (Computer Sci Club) (05/23/84)
I understand the argument that, as a fair coin can produce any
observed distribution, we cannot say on the basis of any
observation that the coin is fair or not fair. Thus any
characterization of the coin as fair or unfair is bound to
be somewhat arbitrary. D. Mitchell seems to indicate that
Bayesian statistics offers a solution to this dilemma but
did not explain this solution clearly. Any comments?
William Hughes