> All IMHO, feedback appreciated.
Dennis,
Frequentist statistics, unlike Bayesian statistics, only allows a hypothesis
to be true or false.
Frequentist test figures quantify the test -- under repeated trials -- rather
than the sample (or the population from which it came). So the p-values indicate
the probability -- if the null hypothesis was true -- of a random sample's
statistics deviating as far as those for your sample.
AFAIK it can say -- within the formalism -- nothing about the probability
of your sample conforming to the null hypothesis. Just statements about the
sampling distribution if the hypothesis is true. Although you might make a
working assumption that the p-value approximates to this.
Berger has some interesting papers on the similarities and differences between
Frequentist and Bayesian approaches -- and the possibility of [partially]
reconciling them.
http://www.stat.duke.edu/~berger/
Peter
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
