[R] LDA with unstandardized discriminant functions

[Samuel Royer Tardif]

Hi,
 
I am currently performing a LDA on soil fatty acid compositions (68 sites, 19 FAs). However, because FAs abundance is caculated in mol% (the molar amount of each FA divided by total molar amount of FAs in the sample), I would like to get unstandardized discriminant functions computed on the original descriptors (see Legendre & Legendre 1998, p.618 for details). How can I do this in R. I've used the function lda(), but the argument scale=FALSE doesn't change anything. Here is an example of my script:
 
#example with 3 variables and 10 sites
x1<-rnorm(10,2,1)
x2<-rnorm(10,4,2)
x3<-rnorm(10,3,2)
X<-cbind(x1,x2,x3)
group<-c("a","a","a","b","b","b","b","c","c","c")
(xlda1<-lda(X,group,scale=TRUE))
(xlda2<-lda(X,group,scale=FALSE))
# It gives the same results
 
Thanks for your help,
 
Samuel Royer Tardif
Ph.D student
Soil ecology
Université de Sherbrooke, Qc, Canada
  		 	   		  
_________________________________________________________________
[[elided Hotmail spam]]