[R] Interaction terms in logistic regression using glm

[Kjetil Halvorsen]

see comments below.

On Wed, Apr 28, 2010 at 4:29 PM, Andrew Miles <rstuff.miles at gmail.com> wrote:
> I recently became aware of the article by Ai and Norton (2003) about how
> interaction terms are problematic in nonlinear regression (such as logistic
> regression).  They offer a correct way of estimating interaction effects and
> their standard errors.
>
> My question is:  Does the glm() function take these corrections into account
> when estimating interaction terms for a logistic regression (i.e. when
> family=binomial)?

No.

  If not, is there a function somewhere that allows for
> correct estimation?

The estimation you get from glm is correct. The discussion in the
paper you referred
is about how to interpret the estimation results! A google search on
the referred paper
(you did'nt give the title), show up various later papers referring to
it, and not supporting their
conclusions.

Linear (and non-linear) model books badly needs chapters with titles
such as "post-estimation analysis". glm does the estimation for you.
It cannot do the analysis for you!

Probably you are looking for something such as CRAN package "effects".

Kjetil



>
> I've looked the documentation for glm and couldn't find an answer, nor have
> I seen the issue addressed in the forums or in the examples of logistic
> regression in R that I've found online.
>
> Thanks!
>
> Andrew Miles
>
> ______________________________________________
> 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.
>