[R] contr.sum, model summaries and `missing' information

[Greg Snow]

Look at the functions dummy.coef, model.tables, and se.contrasts, they may help with what you want.  You can also look at the multcomp package for another approach.

-- 
Gregory (Greg) L. Snow Ph.D.
Statistical Data Center
Intermountain Healthcare
greg.snow at imail.org
801.408.8111


> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of Christophe Rhodes
> Sent: Sunday, September 19, 2010 4:25 AM
> To: r-help at stat.math.ethz.ch
> Subject: Re: [R] contr.sum, model summaries and `missing' information
> 
> Christophe Rhodes <csr21 at cantab.net> writes:
> 
> > I have a dataset with a response variable and multiple factors with
> more
> > than two levels, which I have been fitting using lm() or glm().  In
> > these fits, I am generally more interested in deviations from the
> global
> > mean than I am in comparing to a "control" group, so I use
> contr.sum()
> > as the factor contrasts.  I think I'm happy to interpret the
> > coefficients in the model summary as the effect of a particular
> factor
> > level on the deviation from the overall mean; I'm not after a highly
> > rigorous treatment of these coefficients and their standard errors,
> but
> > rather using them as suggestive of further things to investigate.
> >
> > [...]
> >
> > As far as I can tell, models m1 and m2 are semantically equivalent.
> Is
> > there a straightforward way of extracting the standard error and
> > t-statistic for the `redundant' comparison directly from m1?  I'd
> rather
> > not have to fit two linear models if I can fit just one.
> 
> Did I post this to the wrong list?  I'm still very much interested in
> any answer, or a redirection to a more appropriate forum...
> 
> Thanks,
> 
> Christophe
> 
> ______________________________________________
> 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.