David,
I think that the colorsex approach is the right one, and colorsex should
initially be included as a main effect, because the smooths are centred
for factor by variables (see e.g. ?gam.models). Whether you then choose
to drop this main effect, as it appears to be non-significant, is a
matter of taste (I would tend to leave it in).
best,
Simon
On 27/11/12 16:42, chirleu wrote:> Hi all,
> I wonder if it's possible to include a double interaction in a GAM
formula.
> Example:
>
> If I do this:
> mod=gam(energy~s(size, *by=color, by=sex*, k=5) + temperature, ...)
>
> I get the interaction betwen size*color and size*sex.
>
> But I need size*color*sex, being size a smoother.
> I've created a new variable (colorsex) which combines all the level of
both
> color (2 levels) and sex (2 level), so that I have a new variable with 4
> level. In this case I can do:
>
> mod=gam(energy~s(size, *by=colorsex*, k=5) + temperature, ...)
>
> What do you think of this approach?
>
> In this case, should I also include colorsex (or color*sex) in the
> parametric term *even if it's not significant*(as it's the case)?
>
> Many thanks
>
> David
>
>
>
>
>
>
>
>
> --
> View this message in context:
http://r.789695.n4.nabble.com/interactions-in-GAMs-tp4650987.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> 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.
>
--
Simon Wood, Mathematical Science, University of Bath BA2 7AY UK
+44 (0)1225 386603 http://people.bath.ac.uk/sw283