[R] stepwise selection for conditional logistic regression

[Subha P. T.]

An embedded and charset-unspecified text was scrubbed...
Name: not available
URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20120216/7f64d3b1/attachment.pl>

[Weidong Gu]

stepAIC works for an object of clogit.

Weidong Gu

?Hi,
Is there any function available to do stepwise selection of variables in Conditional(matched) logistic regression( clogit)? step, stepwise ?etc are failing in case of conditional logistic regression. Please help.
Thanks
P.T. Subha
? ? ? ?[[alternative HTML version deleted]]
______________________________________________
R-help at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

[Subha P. T.]

An embedded and charset-unspecified text was scrubbed...
Name: not available
URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20120221/0341b5ea/attachment.pl>

[David Winsemius]

Hi,
Is there any function available to do stepwise selection of
variables in Conditional(matched) logistic regression( clogit)?
step, stepwise etc are failing in case of conditional logistic
regression.

"Failing" is open to a variety of interpretation. Can you offer an
example and describe what you mean or quote an error message?

I tested the advice in this posting from C. Berry with the example in ?
clogit and got appropriate results:

http://finzi.psych.upenn.edu/Rhelp10/2010-January/226165.html

Please help.

Stepwise procedures are supported somewhat grudgingly on r-help. You
ought to read some of the critical comments about stepwise procedures
in the Archives:

http://search.r-project.org/cgi-bin/namazu.cgi?query=stepwise+significance&max=100&result=normal&sort=score&idxname=functions&idxname=Rhelp08&idxname=Rhelp10&idxname=Rhelp02

Another search strategy might be "stepwise Harrell".

[Steve Lianoglou]

Also, when you're doing reading through David's suggestions:

On Fri, Feb 17, 2012 at 10:41 AM, David Winsemius
<dwinsemius at comcast.net> wrote:
[snip]

Stepwise procedures are supported somewhat grudgingly on r-help. You ought
to read some of the critical comments about stepwise procedures in the
http://search.r-project.org/cgi-bin/namazu.cgi?query=stepwise+significance&max=100&result=normal&sort=score&idxname=functions&idxname=Rhelp08&idxname=Rhelp10&idxname=Rhelp02
Another search strategy might be "stepwise Harrell".

Just keep in the back of your mind somewhere that the glmnet library
can fit "GLMs via penalized maximum likelihood [and] ... Fits linear,
logistic and multinomial, poisson, and Cox regression models" (taken
from `?glmnet`) over a grid of lambda params for you.

HTH,
-steve

[Subha P. T.]

An embedded and charset-unspecified text was scrubbed...
Name: not available
URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20120221/70be1861/attachment.pl>

[Subha P. T.]

An embedded and charset-unspecified text was scrubbed...
Name: not available
URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20120221/4ed55541/attachment.pl>

[David Winsemius]

Hi David
My data set has about 20 significant variables and step function
with logistic regression is working fine( in R-commander). I tried
to get conditional logistic by introducing the stratum variable and
clogit. The clogit is not converging but is giving the summary of
the model. When step/step AIC/..are used, the message given is
"ERROR: number of rows in use has changed".

Caveat: I do not generally use stepwise methods and I have no
experience with this particular message. Nonetheless it suggests that
you have some variable in the model tat differs from the others with
respect to missingness. Have you tried using subset() or
complete.cases() to select a set of non-missing data for all tested
variables?

--
David Winsemius, MD
West Hartford, CT

[Frank Harrell]

Stepwise variable selection is an invalid statistical method. Who or which
book recommended it?
Frank

?Hi,
Is there any function available to do stepwise selection of variables in
Conditional(matched) logistic regression( clogit)? step, stepwise ?etc are
failing in case of conditional logistic regression. Please help.?
Thanks
P.T. Subha
[[alternative HTML version deleted]]
______________________________________________
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

-----
Frank Harrell
Department of Biostatistics, Vanderbilt University
--
View this message in context: http://r.789695.n4.nabble.com/stepwise-selection-for-conditional-logistic-regression-tp4396607p4410260.html
Sent from the R help mailing list archive at Nabble.com.