[R] repost my question of cda in case

[Spencer Graves]

      For linear discriminate analysis with only 2 classes, you get one 
new variable, that being a straight line between the means of the two 
groups. 

      With 3 classes, the 3 points determine a plane defined by 2 lines 
or "new functions" = linear combinations of the original variables.  
However, if the distinctions are not statistically significant, you will 
get less than 2. 

      hope this helps. 
      spencer graves

WeiWei Shi wrote:

>Dear R-helpers:
>
>I sent this question 3 days ago but I didn't get any reply. In case
>this question was somewhat not seen by people who happpened to know
>the answer, I repost it here. Sorry for bother but I am kind of
>needing some help. BTW, if the question itself was not well expressed,
>please let me know.
>
>The question is as followed:
>
>
>I am wondering if I can get some general help or source about
>canonical discriminant analysis in R.
>
>My idea is trying to linearly "combine" 300 variables supervisely
>(according to the class lables to the observations". I think it is
>kinda PCA to do some decreasing dimentionality work, but w/
>considering the class and I used SAS to do CDA proc before.
>
>But I read the introduction from sas on this proc and found the
>following statement:
>
>"The process of extracting canonical variables can be repeated until
>the number of canonical variables equals the number of original
>variables or the number of classes minus one, whichever is smaller.
>"
>does it mean I can only have two new variables if  I only have 2 classes?
>
>I am not a stat guy and sorry for the question if it should not be
>addressed here.
>
>Ed
>
>______________________________________________
>R-help at stat.math.ethz.ch mailing list
>https://stat.ethz.ch/mailman/listinfo/r-help
>PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
>  
>