On Thu, 23 Mar 2000 08:55:00 GMT, [EMAIL PROTECTED] wrote:
> Hi all,
> This is a question on Log-Reg methodology.
> How can I evaluate the individual influence of each variable in the
> prediction of the logistic regression?.
- this is almost identical to considering OLS regression. You are
looking at "partial" regression coefficients when you have the whole
equation.
The logistic solution has potentially *two* tests for contribution of
the single variable -- the curvature of the likelihood space (second
derivative) might not give exactly the same answer as the change in
the Log-Likelihood chi-squared when deleting the variable. The thing
to note about that is whether it happens: if the tests differ, it
indicates some instability or non-robustness in the solution.
> I usually perform a logistic regression with forward selection so, I
> have the order of variables entered in the model.
> Now, to compare the individual influence, which statistic should I
> use? How correlation between variables is taken in account? Is it
> correct to perform individual log-reg
> for each variable and then compare the predictions?
Doing step-wise entry of variables for logistic is just as bad an idea
as it is for OLS regression. Especially relevant for small N: Note
that logistic is even *worse* than OLS about capitalizing on chance.
You can read about stepwise, and references, in my stats-FAQ.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===========================================================================
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Please DO NOT COMPLAIN TO
THE POSTMASTER about these messages because the postmaster has no
way of controlling them, and excessive complaints will result in
termination of the list.
For information about this list, including information about the
problem of inappropriate messages and information about how to
unsubscribe, please see the web page at
http://jse.stat.ncsu.edu/
===========================================================================
