[R] GLM problem

[Prof Brian Ripley]

On Wed, 30 Jun 2004, Varrin muriel wrote:

> HI, I am a studient, so don't be surprise if my question seems so simple
> for you...

> I have a dataframe with 6 qualitative variables divided in 33
> modalities, 2 qualitatives variables and 78 lines. I use a glm to know
> wich variables have interactions... I would like to know if its normal
> that one (the first in alphabetical order) of the modalities of each
> qualitatives variable doesn't appear in the results?

Yes.  That's the way R codes variables.  I'm not going to rewrite the
literature here, so please consult a good book (and naturally I will
recommend Venables & Ripley, 2002).  Others may be able to tell you if any
of the French guides on CRAN or elsewhere address this point (Emmanuel
Paradis's seems not to get that far).

> Here is what i did:
> glm<-glm(data$IP~data$temp*data$fvent*data$dirvent*data$Ensol+data$mois*data$mil, 
> family=gaussian)

Please use something like

glm(IP ~ temp+fvent, data=data, family=gaussian)

and don't call the result `glm' (or call your data, `data').  Also, this
is not a question about GLMs with that family but about linear models, so
why not use lm.

> summary(glm)
> anova(glm, test="Chisq")
> thanck you a lot and excuse me again for my little and perhaps simple question
> muriel varrin (french student in biostatistical)

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595