🚀 go-pugleaf

Dear Stat-L Colleagues,

Any classification enthusiasts reading the list?  I would like to
discuss results from a comparative study of logistic regression and
discriminant analysis with such a person.

We recently tested several variants of the above methods on 5 real data
sets of various types, including discrete data.  The results suggest
that the comparative merits of the methods depend almost entirely on the
priors used in the discriminant analysis (and the composition of the
training set for logistic regession, this serving as a type of prior).
In retrospect this result doesn't seem al all surprising, but I wasn't
expecting it.  Is this a well known phenomenon?  Certainly the whole
topic of choosing the priors is barely mentioned in most multivariate
statistics texts and in the literature I have surveyed about these two
methods.

I will appreciate all comments.  Please reply directly to me, & I will
summarize responses for the list if there are many.

...Mike
*====================================================*
*  Michael H. Prager                                 *
*  U.S. National Marine Fisheries Service            *
*  Southeast Fisheries Science Center                *
*  75 Virginia Beach Drive                           *
*  Miami, Florida  33149  USA                        *
*  Internet:    mprager%pontus.span@umigw.miami.edu  *
*  Telephone:   305-361-4218  (FTS 350-1218)         *
*====================================================*