manipulating PCA output in R

I am doing a PCA on a single dataset and want to control which variables
are plotted as vectors on the biplot produced. How can I first determine
which variables are most important in producing the ordination and how do
I then control what variables are included as vectors in the biplot?

I'm not sure what you mean by "most important", but I think you want to know
the loadings on the components, which you can retrieve easily from the
object returned by princomp or prcomp (it depends on which function you
used). See ?prcomp and ?princomp. With prcomp the rotations are printed by
default. With princomp, use loadings(princomp.object).

Not sure about suppressing certain arrows (variables) from the plot.

-----
David Hewitt
Research Fishery Biologist
USGS Klamath Falls Field Station (USA)

View this message in context: http://www.nabble.com/manipulating-PCA-output-in-R-tp17877744p17916947.html
Sent from the r-sig-ecology mailing list archive at Nabble.com.

An embedded and charset-unspecified text was scrubbed...
Name: not available
URL: <https://stat.ethz.ch/pipermail/r-sig-ecology/attachments/20080617/d04b35d5/attachment.pl>

Quoting David Hewitt <dhewitt37 at gmail.com>:

I am doing a PCA on a single dataset and want to control which variables
are plotted as vectors on the biplot produced. How can I first determine
which variables are most important in producing the ordination and how do
I then control what variables are included as vectors in the biplot?

I'm not sure what you mean by "most important", but I think you want to know
the loadings on the components, which you can retrieve easily from the
object returned by princomp or prcomp (it depends on which function you
used). See ?prcomp and ?princomp. With prcomp the rotations are printed by
default. With princomp, use loadings(princomp.object).

Not sure about suppressing certain arrows (variables) from the plot.

The question was vague, but I assume it concerned prcomp (or princomp)  
and biplot.prcomp (or biplot.princomp). Selecting variables is  
particularly easy in these functions, because they take one-sided  
model formulae. To pick vars x3, x7 and x9 from a data frame df you  
just write prcomp(~ x3 + x7 + x9, data=df). However, there seems not  
be an easy way of omitting variables used in prcomp (or princomp) in  
biplot functions (there may be this option if you use rda function in  
prcomp but not for arrows, only for points or text, but I don't care  
to check the documentation). Since you ask this, I think you'd rather  
want to have factor analysis than PCA. It may be good to consult a  
statistician or a book on this issues.

Best wishes, Jari Oksanen