>If psi precognition were operating, one would expect a significantly
>elevated (or perhaps a significantly depressed) vote for the six numbers
>popping out of the randomiser. I didn't find that overall, to my chagrin -
>but I did locate a very weird effect in which the *most drastically deviant
>positive residuals correlated with targets to a far greater extent than one
>would expect by chance*. Opinions differ over whether this post hoc
>finding supports the psi hypothesis or is an artefact. I think that when
>the suggested artefact is factored in, there is still evidence for
>precognition in the data.
Post-hoc comparisons have essentially no value for proving things; only for
suggesting hypotheses to be test. I speak from experience; I worked in
medical devices for 7 years doing statistics. I've seen post-hoc results
in our studies as far out as 1 in 20,000 turn out to be completely bogus.
Post-hoc comparisons are most definitely a case where enough monkeys
will always produce something.