[R] using MANOVA in R

[Prof Brian D Ripley]

On Tue, 9 May 2000, Torsten Hothorn wrote:

> On 9 May 2000, Peter Dalgaard BSA wrote:
> 
> > Prof Brian D Ripley <ripley at stats.ox.ac.uk> writes:
> > 
> > > Not really any of it!  multilm fits multivariate linear models and does
> > > Hotelling T^2 tests. It does not group terms for AOV, and I think is only
> > > applicable for continuous variates (not factor explantory variables).
> > > Certainly the help and examples only discuss that case.
> > 
> > Had a closer look. The factor expl.var. would seem to be there
> > case-wise, so you can build the models, but Wilk's Lambda and friends
> > are absent, so you cannot (easily) do multi-df model reduction tests.
> > However, it's no too far off. Is the author planning to develop it
> > further?
> 
> Now it is my turn, I think :-)
> 
> 1) "multilm" is on my list, but there are some theoretical problems
> (e.g. is the use of Moore-Penrose numerically appropriate ?) Any help by
> other distributors would be very welcome!
> 
> 2) "multilm" was written for the use of the stabilized multivariate tests
> by Laeuter and Kropf, T^2 is only a by-product. Therefore the
> documentation is heavily biased on this. 
> 
> 3) I do think that one can take factors as design variables (there is an
> Iris example, I think). 

One can fit the model, but as the Iris example shows, it assumes three
continuous vars, NOT one factor in a model formula, and so the information
given is not appropriate for a manova as I understand that term.
Manova includes a series of tests of submodels.

Here's an S example of manova:

       wafer.manova <- manova(cbind(pre.mean, post.mean) ~ maskdim +
          visc.tem + spinsp, wafer)

       summary(wafer.manova) # manova table with Pillai's trace

          Df Pillai Trace approx. F num df    den df    P-value 
  maskdim 1  0.42662      4.09229   2         11        0.04693
 visc.tem 2  0.5809       2.45605   4         24        0.07305
   spinsp 2  0.47904      1.88978   4         24        0.14495
Residuals 12                                                   

so factors in the model are grouped and several submodels have been
fitted.

-- 
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 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._