[R] Discriminant plot

[Alain Guillet]

Hello Alejo,

Please, keep sending your post to the R-help mailing list in order other 
people can also answer.

The type of lda_analysis is lda and that is normal and it also is 
perfectly normal to find a different type for predict(lda_analysis)$x. 
Moreover the example of the lda() function about iris gives me the exact 
same types for the object z (of the example) and for predict(z).

When you plot lda_analysis, you use the function plot.lda whereas you 
use the function plot for the predict object.

As I told you in my previous e-mail the predicted class are not the 
class of X$G3 so it is normal if the two plots are not exactly the same.
which(predict(lda_analysis)$class != X$G3) gives you all the 
observations that are predicted in a different category from X$G3. Look 
at this points and you can see they are the only different points from 
the two plots (the coordinates are the same).

Alain


Alejo C.S. wrote:
> Hi Alain,
>
> I thought  (in the worng way I see)  that the predict function applied 
> to an object of class "lda" returned the coordinates of the 
> discriminant axes. When doing the same to iris data, the original 
> classes are the same than those returned by predict. Is not the case 
> with my data, if you compare the original classes with those returned 
> by predict(), the are different.
> I'm really confused now.......
>
> Regards,
>
>
> Alejo
>
> 2009/10/15, Alain Guillet <alain.guillet at uclouvain.be 
> <mailto:alain.guillet at uclouvain.be>>:
>
>     Hi Alejo,
>
>     According to my knowledge the two plots are different because in the
>     first one a point belongs to a group depending on its group in the
>     data
>     whereas in the second plot a point belongs to the group predicted
>     by the
>     linear discriminant analysis.
>
>     I hope somebody will correct me if I am wrong.
>
>     Alain
>
>
>     Alejo C.S. wrote:
>
>         Hi Alain, this is the code:
>
>
>         library(MASS)
>         library(mda)
>
>
>         #data attached, first column "G3" group membership
>
>         X <- read.table("data", header=T)
>
>         lda_analysis <- lda(formula(X), data=X)
>
>         plot(lda_analysis, col=palette()[X$G3])
>
>         #the above plot is completely different to:
>
>         plot(predict(lda_analysis)$x, type="n")
>         text(predict(lda_analysis)$x,
>         labels=predict(lda_analysis)$class,
>         col=palette()[predict(lda_analysis)$class])
>
>         The above code only reproduce the first plot using predict to
>         obtain coordinates and classes for the first tow discriminant
>         axis.
>
>         Thanks ,
>
>         Alejo
>
>
>     -- 
>     Alain Guillet
>     Statistician and Computer Scientist
>
>     SMCS - Institut de statistique - Université catholique de Louvain
>     Bureau c.316
>     Voie du Roman Pays, 20
>     B-1348 Louvain-la-Neuve
>     Belgium
>
>     tel: +32 10 47 30 50
>
>
>

-- 
Alain Guillet
Statistician and Computer Scientist

SMCS - Institut de statistique - Université catholique de Louvain
Bureau c.316
Voie du Roman Pays, 20
B-1348 Louvain-la-Neuve
Belgium

tel: +32 10 47 30 50