Some myths concerning statistical hypothesis testing
Glen M. Sizemore
gmsizemore2 at yahoo.com
Thu Nov 21 15:19:14 EST 2002
Only a small portion of this thread has been about Bayesian statistics. And
incidentally, there is a whole different approach in which one establishes a
baseline for some preparation (maintains the preparation under fixed
conditions until the possible range of variation is directly observed) and
then introduces a variable or changes the level of a variable. Any effect is
judged by comparison with the range of variation observed under baseline
conditions. This allows for the direct demonstration of within- and
across-preparation reliability, and produces functions that are relevant to
individual preparations (rather than being relevant to "sample averages,"
which may have little to do with the phenomena as they exist in the
individual. Such a view eschews averaging across preparations.
"fredm" <Spam_not_fredm at frontiernet.net> wrote in message
news:utqd7ado4sko56 at corp.supernews.com...
> I'd be interested in links/books as well. One that I found is:
> http://astrosun.tn.cornell.edu/staff/loredo/bayes/tjl.html go to the link
> From Laplace to Supernova SN 1987A: Bayesian Inference in Astrophysics
> which is:
> It is pretty good at explaining (the only?) two competing ideas of
> probability and why the author elects to use the Bayesian methods.
> "B Traven" <bbmeme at hotmail.com> wrote in message
> news:682a6547.0211210028.158d4e8d at posting.google.com...
> > Does anyone have a citation (or better yet, link), to an overview of
> > the various schools of thought on probability, statistics, and
> > inference? Something that describes how, where, and why schools of
> > thought differ.
> > - Bb
> > robert_dodier at yahoo.com (Robert Dodier) wrote in message
> > >
> > > Just to clarify the grounds of the debate, the questions of interest
> > > are not mathematical in nature -- so far as I know all parties agree
> > > on the theorems of probability, measure theory, etc. and nobody claims
> > > that their opponents have a false derivation or some error like that.
> > >
> > > The debate is best characterized as a scientific in nature --
> > > specifically, there is disagreement as to what the A, B, C, and
> > > X, Y, Z in the equations can stand for. It is something like a
> > > physicist exhibiting an equation for balance of phlogiston --
> > > even if the equation itself is OK, some people will object to
> > > interpreting the quantity P as a massless fluid that transfers energy.
> > >
> > > Specifically, in the case of statistics, one group claims that it is
> > > meaningful to assess probability for any proposition, be it concerning
> > > random variables or otherwise. Another influential group claims that
> > > is incorrect, and some mode of reasoning other than probability is
> > > required for any proposition not concerning random variables. This
> > > disagreement as to the scope of probability has lead to vastly
> > > different methodologies, and never the twain shall meet, AFAICT.
> > >
> > > Statistics courses for non-majors are almost entirely taught by
> > > the "probability for random variables only" party; this is a
> > > historical and sociological phenomenon. OTOH, I am aware that the
> > > other persuasion is popular in many computer science departments,
> > > specifically as it makes automated reasoning much easier to formulate.
> > >
> > > For what it's worth,
> > > Robert Dodier
More information about the Neur-sci