biplot.princomp - changing score labels

Dear R helpers,

When producing a PCA biplot, vectors of environmental variables (as 
red arrows with labels) and scores of the observations (black labels 
(observation names)) are plotted by default. How can I change the 
graphical output? Let's say I would like that the scores are plottet 
only as symbols and not text. The only solution I found was this post 
in the help archive
http://tolstoy.newcastle.edu.au/R/e2/help/06/12/7298.html
suggesting "cheating" like:

biplot(test.pca, xlabs=c("0","0","0","0","0","0...

and using letters instead of symbols.
This works somehow, but I am still not able the change e.g. the colour 
or size of only the scores. Whatever I try to do with "col=...", 
"cex=..." and so on changes either all labels in the plot or none. Can 
someone tell me what is the trick to only modify either the 
environmental variables or the scores? Also e.g. the type of the arrow?

Thanks in advance,
Axel


------------------------------------------------------------------------

OK, the one thing I figured out:
Is should be like:
biplot(test.pca, cex=c(2,1), col=c("red","green")...
to change size, colours etc separately. But I still don't know how change 
lables of observations to symbols properly.

Tipps? Thanks again,
Axel

Dear R helpers,

When producing a PCA biplot, vectors of environmental variables (as red 
arrows with labels) and scores of the observations (black labels 
(observation names)) are plotted by default. How can I change the graphical 
output? Let's say I would like that the scores are plottet only as symbols 
and not text. The only solution I found was this post in the help archive
http://tolstoy.newcastle.edu.au/R/e2/help/06/12/7298.html
suggesting "cheating" like:

biplot(test.pca, xlabs=c("0","0","0","0","0","0...

and using letters instead of symbols.
This works somehow, but I am still not able the change e.g. the colour or 
size of only the scores. Whatever I try to do with "col=...", "cex=..." and 
so on changes either all labels in the plot or none. Can someone tell me 
what is the trick to only modify either the environmental variables or the 
scores? Also e.g. the type of the arrow?

Thanks in advance,
Axel

On Tue, 24 Feb 2009, Axel Strau? wrote:

OK, the one thing I figured out:
Is should be like:
biplot(test.pca, cex=c(2,1), col=c("red","green")...
to change size, colours etc separately. But I still don't know how 
change lables of observations to symbols properly.

That's not part of the design of the function, so just make a copy and 
edit to meet your fancies.

The designer of biplot.princomp.

Prof Brian Ripley schrieb:

On Tue, 24 Feb 2009, Axel Strau? wrote:

OK, the one thing I figured out:
Is should be like:
biplot(test.pca, cex=c(2,1), col=c("red","green")...
to change size, colours etc separately. But I still don't know how change
lables of observations to symbols properly.

That's not part of the design of the function, so just make a copy and
edit to meet your fancies.

The designer of biplot.princomp.

The idea behind my question was actually not a "styling one" but to provide
additional information in the graph. My observations have different species
richness and I wanted to adopt symbol size to the number of species to show
the change of species richness along the PCs. Anyway, thanks for the comment
- and for designing biplot.princomp.

Axel

Perhaps this may help you.

Regards




data(iris3)

ir <- rbind(iris3[ , , 1], iris3[ , , 2], iris3[ , , 3])
    
  ir.pca <- princomp(ir)
  biplot(ir.pca)

# Rehacer Biplot
# Calcular Factor para re-escalar scores y eigenvectores
  lambda <- ir.pca$sdev[1:2] * sqrt(ir.pca$n.obs)

  scores <- t( t(ir.pca$scores[ , c(1, 2)]) / lambda 
)                        
  variables <- t( t(ir.pca$loadings[ , c(1, 2)]) * lambda )
 
     
  x11()
  plot(scores, type="p", xlim=c(-0.22, 0.24), ylim=c(-0.2, 0.2),
  pch=c(rep(16, 50), rep(21, 50), rep(14, 50)),
  col=c(rep("red", 50), rep("blue", 50), rep("green", 50)))
    abline(v=0, h=0, lty=3)
  par(new=TRUE)
  plot(variables, type="n", xaxt="n", yaxt="n", xlim=c(-22, 24), 
ylim=c(-20, 20))
    arrows(0, 0, variables[ , 1], variables[ , 2], len=0.1, col="red")
    text(2*variables, rownames(variables), col="red", xpd=TRUE)
    axis(3); axis(4)



On 25 Feb 2009 at 9:52, Axel Strau? wrote:

Date sent:                  Wed, 25 Feb 2009 09:52:54 +0100
From:                         Axel Strau? <a.strauss at tu-bs.de>
To:                            R-help at r-project.org
Subject:                     Re: [R] biplot.princomp - changing score 
labels

Prof Brian Ripley schrieb:

On Tue, 24 Feb 2009, Axel Strau? wrote:

OK, the one thing I figured out:
Is should be like:
biplot(test.pca, cex=c(2,1), col=c("red","green")...
to change size, colours etc separately. But I still don't know how
change lables of observations to symbols properly.

That's not part of the design of the function, so just make a copy
and edit to meet your fancies.

The designer of biplot.princomp.

The idea behind my question was actually not a "styling one" but to
provide additional information in the graph. My observations have
different species richness and I wanted to adopt symbol size to the
number of species to show the change of species richness along the
PCs. Anyway, thanks for the comment - and for designing
biplot.princomp.

Axel

-- 
H?ctor Villalobos <hvillalo at ipn.mx>
 CICIMAR - IPN
 A.P. 592. Col. Centro
 La Paz, Baja California Sur, M?XICO. 23000
 Tels. (+52 612) 122 53 44; 123 46 58; 123 47 34  ext. 82425
 Fax.  (+52 612) 122 53 22