[R] Multiple dependent variables

[Peter Dalgaard]

Simon CHAMAILLE <s.chamaille at wanadoo.fr> writes:

> Dear friends,
> I'm stuck with the following problem:
> I would like to do a multiple regression with muliple dependent variables, and i don't know how to write my formula in the model. Someone in previous messages offered the cbind method, but the result is just as many regression as the number of dependent variables, it is just a time saving method; i'm looking for a method closer to a MANOVA, but without factors. well, a standard multiple rgression with multiple dependent variable, i.e. looking for variables that are significant in all the dependent variables.
> If one of you have an idea, I welcome it.
> and happy new year to all
> simon

Some of this stuff appears to be not quite implemented, but have a
look at

> summary(manova(cbind(y1,y2)~x1+x2))
          Df   Pillai approx F num Df den Df Pr(>F)
x1         1 0.079014 0.257378      2      6 0.7812
x2         1 0.006782 0.020485      2      6 0.9798
Residuals  7
> summary(manova(cbind(y1,y2)~x2+x1))
          Df   Pillai approx F num Df den Df Pr(>F)
x2         1 0.006930 0.020936      2      6 0.9794
x1         1 0.078886 0.256927      2      6 0.7815

The annoying thing is that drop1 and friends are not there, so you
have to arrange terms so that the sequential tests make sense.
However, all the bits are clearly there.

Another annoying thing is that manova() is not analogous to anova().

-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907