R help - Nov 2010 - help! kennard-stone algorithm in soil.spec packages does not work for my dataset!!!

http://r.789695.n4.nabble.com/file/n3031344/RSV.Rdata RSV.Rdata 
I want to split my dataset to training set and test set using
kennard-stone(KS) algorithm, it is lucky there is R packages soil.spec to
implement it.

but when I used it to my dataset, it does not work, who can help me, how
reasons is it, below, it is my code, and my data in the attachment.

ks<-ken.sto(x,per="TRUE",per.n=0.3,va="FALSE",sav="FALSE")
ks

% results

$`Chosen sample names`
NULL

$`Chosen row number`
integer(0)

$`Chosen calibration sample names`
[1] "NULL"

$`Chosen calibration row number`
[1] "NULL"

$`Chosen validation sample names`
[1] "NULL"

$`Chosen validation row number`
[1] "NULL"

attr(,"class")

why it is all NULL ?


and
>
ks<-ken.sto(x,per="TRUE",per.n=0.3,va="TRUE",sav="FALSE")
Error in val.min[i] <- blub[sample(length(blub), 1)] : 
  replacement has length zero
In addition: Warning message:
In min(prco[-cal.start.n, i]) :
  no non-missing arguments to min; returning Inf

if I set va="TRUE", appearing the errors.

I hope some friends can help me !
-- 
View this message in context:
http://r.789695.n4.nabble.com/help-kennard-stone-algorithm-in-soil-spec-packages-does-not-work-for-my-dataset-tp3031344p3031344.html
Sent from the R help mailing list archive at Nabble.com.

On Nov 7, 2010, at 6:53 PM, bbslover wrote:
>
> http://r.789695.n4.nabble.com/file/n3031344/RSV.Rdata RSV.Rdata
> I want to split my dataset to training set and test set using
> kennard-stone(KS) algorithm, it is lucky there is R packages  
> soil.spec to
> implement it.
>
> but when I used it to my dataset, it does not work, who can help me,  
> how
> reasons is it, below, it is my code, and my data in the attachment.
No attachment, athough partial credit for making the effort. Most  
probable cause... failing to read Posting Guide.

-- 
David.
>
>
ks<-ken.sto(x,per="TRUE",per.n=0.3,va="FALSE",sav="FALSE")
> ks
>
> % results
>
> $`Chosen sample names`
> NULL
>
> $`Chosen row number`
> integer(0)
>
> $`Chosen calibration sample names`
> [1] "NULL"
>
> $`Chosen calibration row number`
> [1] "NULL"
>
> $`Chosen validation sample names`
> [1] "NULL"
>
> $`Chosen validation row number`
> [1] "NULL"
>
> attr(,"class")
>
> why it is all NULL ?
>
>
> and
>
>>
ks<-ken.sto(x,per="TRUE",per.n=0.3,va="TRUE",sav="FALSE")
> Error in val.min[i] <- blub[sample(length(blub), 1)] :
>  replacement has length zero
> In addition: Warning message:
> In min(prco[-cal.start.n, i]) :
>  no non-missing arguments to min; returning Inf
>
> if I set va="TRUE", appearing the errors.
David Winsemius, MD
West Hartford, CT

http://r.789695.n4.nabble.com/file/n3032045/rsv1.txt rsv1.txt 


I am very grateful to David's suggestion, here , I upload my dataset
"rsv1.txt", also the question, 

ks<-ken.sto(rsv1,per="TRUE",per.n=0.3,va="FALSE",sav="FALSE")

it does not work, all results are NULL, i do not known why it is ?

hope, friends can give me a hand!

thanks 

kevin

-- 
View this message in context:
http://r.789695.n4.nabble.com/help-kennard-stone-algorithm-in-soil-spec-packages-does-not-work-for-my-dataset-tp3031344p3032045.html
Sent from the R help mailing list archive at Nabble.com.

Hi all

Certainly there are some problems with the "ken.sto" function. In
addition
there are some considerations that need to be taken into account before
using this code. For instance, it projects the data onto a principal
component space prior sampling. In this case is necessary to check if a PCA
is really necessary. Second, this function uses the Euclidean distance (ED)
as metric for selecting the samples. By using the ED directly on the PCs,
the PCs with high explained variance will ?dominate? the measurements. In
this case is better to standardize the PCs prior ED computation or use the
Mahalanobis distance.
Whatever, I have modified the ?ken.sto? code and now seems to be that it
works correctly. If some of you still need it just drop me an e-mail.
Regards, 

Leonardo Ramirez-Lopez, 

Researcher
Georges Lema?tre Centre for Earth and Climate Research
Earth and Life Institute
Universit? Catholique de Louvain,
3 Place Louis Pasteur
1348, Louvain la Neuve
Belgium
e-mail: leonardo.ramirez at uclouvain.be

http://www.uclouvain.be/en-teclim.html


Ph.D Researcher
Physical Geography and Soil Science
Institute of Geography
University of T?bingen
R?melinstr. 19-23
72070 T?bingen
Germany
http://www.geographie.uni-tuebingen.de/ 




--
View this message in context:
http://r.789695.n4.nabble.com/help-kennard-stone-algorithm-in-soil-spec-packages-does-not-work-for-my-dataset-tp3031344p4306703.html
Sent from the R help mailing list archive at Nabble.com.