R help - Apr 2011 - Orthoblique rotation on eigenvectors (SAS VARCLUS)

Hi All,

I'd like to build a package for the community that replicates the output
produced by SAS "proc varclus". According to the SAS documentation,
the
first few steps are:

1. Find the first two principal components.
2. Perform an orthoblique rotation (quartimax rotation) on eigenvectors.
3. Assign each variable to the rotated component with which it has the
higher
   squared correlation.
The cartoon example below attempt to do this, but I found my results differ
from SAS in "pc3" (i.e,  the standardize component scores).

I'd appreciate your help in whether you see anything wrong in
"pc2" or
"pc3"?

set.seed(1)
x1=rnorm(200); x2=0.9*x1; x3=0.7*x1; x4=x1*x1; x5=x1*x1*x1;
x6=rnorm(200); x7=0.9*x6; x8=0.7*x6; x9=x6*x6; x10=x6*x6*x6;
x <- cbind(x1,x2,x3,x4,x5,x6,x7,x8,x9,x10)

require(GPArotation)
pc1 <- princomp(x, cor = TRUE, scores = TRUE)
pc2 <- quartimax(pc1$loadings[,1:2],normalize=TRUE)$loadings
pc3 <- scale(x%*% pc2)
pc4 <- apply(x, 2, function(x) cor(x, pc3)^2)

Thanks in advance for any help!

Axel.

	[[alternative HTML version deleted]]

See the varclus function in the Hmisc package.
Frank


Axel Urbiz wrote:
> 
> Hi All,
> 
> I'd like to build a package for the community that replicates the
output
> produced by SAS "proc varclus". According to the SAS
documentation, the
> first few steps are:
> 
> 1. Find the first two principal components.
> 2. Perform an orthoblique rotation (quartimax rotation) on eigenvectors.
> 3. Assign each variable to the rotated component with which it has the
> higher
>    squared correlation.
> The cartoon example below attempt to do this, but I found my results
> differ
> from SAS in "pc3" (i.e,  the standardize component scores).
> 
> I'd appreciate your help in whether you see anything wrong in
"pc2" or
> "pc3"?
> 
> set.seed(1)
> x1=rnorm(200); x2=0.9*x1; x3=0.7*x1; x4=x1*x1; x5=x1*x1*x1;
> x6=rnorm(200); x7=0.9*x6; x8=0.7*x6; x9=x6*x6; x10=x6*x6*x6;
> x <- cbind(x1,x2,x3,x4,x5,x6,x7,x8,x9,x10)
> 
> require(GPArotation)
> pc1 <- princomp(x, cor = TRUE, scores = TRUE)
> pc2 <- quartimax(pc1$loadings[,1:2],normalize=TRUE)$loadings
> pc3 <- scale(x%*% pc2)
> pc4 <- apply(x, 2, function(x) cor(x, pc3)^2)
> 
> Thanks in advance for any help!
> 
> Axel.
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> 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.
> 

-----
Frank Harrell
Department of Biostatistics, Vanderbilt University
--
View this message in context:
http://r.789695.n4.nabble.com/Orthoblique-rotation-on-eigenvectors-SAS-VARCLUS-tp3438695p3440064.html
Sent from the R help mailing list archive at Nabble.com.

I really like your varclus function Frank.  I have a question about the
quartimax rotation of the 1st two eigen vectors as the SAS VARCLUS algorithm
dictates.  According to the SAS user guide example of VARCLUS the only thing
needed to perform the cluster analysis is the variable correlation matrix. 
Is that the case?  The following is the SAS user guide example website.

http://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_varclus_sect020.htm

Larry



--
View this message in context:
http://r.789695.n4.nabble.com/Orthoblique-rotation-on-eigenvectors-SAS-VARCLUS-tp3438695p4187270.html
Sent from the R help mailing list archive at Nabble.com.