[EMAIL PROTECTED] (wuzzy) wrote, in part:

> What I was trying to do (probably failed) is a bayesian logistic
> regression.

OK, thanks for the clarification; I didn't understand what was
the goal before.

It is well known that the parameters for a logistic regression
may not be well defined by the available training data. 
The right thing to do, from a Bayesian p.o.v., is to average
output from the logistic regression model over the posterior
distribution of the parameters. It can be this simple:
a <- (samples from posterior); y <- (logistic regression output
for given input and parameters a); average(y). If you bear this
is mind, I think it will make the task comprehensible.

There is a worked example involving logistic regression in 
Bayesian Data Analysis, by Gelman, Carlin, Stern, and Rubin,
somewhere around page 82 (first edition).

For what it's worth,
Robert Dodier
--
Far better an approximate answer to the right question, which is often
vague, than an exact answer to the wrong question, which can always be
made precise. -- John W. Tukey
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to