Dear R community and Kurt,
I am using consensus.r from clue to compute hard consensus partitions
(method: "hard/euclidean"; same considerations apply for "hard/manhattan").
The function gives in output a hard consensus partition that minimizes a
criterion measure Crit, namely the sum of the weighted dissimilarities
between the input partitions on the one hand, and the consensus
partition on the other.
I would like to have an exact value for Crit (consensus.r only shows
approximate values when verbose=TRUE).
To this purpose, I manually compute Crit using the cl_dissimilarity
function from the same package.
The problem I have is that the Crit I compute manually differs widely
from that computed by criterion.r (compare "Minimum" with "Manually
computed criterion" after running the sample code).
I figure the difference arises because my criterion measure does not
weight the cases.
However, I wonder:
1. why does the consensus partition search algorithm have to weight the
cases? Is the weighting part of the heuristic?
2. how can I recover these weights (hacking consensus.r is not really
straightforward, since it includes many subfunctions)?
3. is there something wrong in my R code (pasted below)?
Thank you for any suggestion, and for this very helpful package.
Best,
Bruno
p.s. in the code, you can set nnruns to 1 if you do not want to load
your machine.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Bruno L. Giordano, Ph.D.
Music Perception and Cognition Laboratory
CIRMMT http://www.cirmmt.mcgill.ca/
Schulich School of Music, McGill University
555 Sherbrooke Street West
Montréal, QC H3A 1E3
Canada
Office: +1 514 398 4535 ext. 00900
http://www.music.mcgill.ca/~bruno/
###################### CODE #############################
library(clue)
##### partitions
p1=c(1,2,3,4,5,6,7,8,9,1,1,2,2,3,3,4,4,1,1,2,2,3,3,4,4)
p2=c(1,1,2,2,3,3,4,4,1,1,2,2,3,3,4,4,1,1,2,2,3,3,4,4,1)
p3=c(1,1,2,2,3,3,4,4,1,1,2,2,3,3,4,4,1,1,2,2,3,3,4,5,5)
##### partitions ensemble
p=as.cl_ensemble(list(as.cl_partition(p1),as.cl_partition(p2),as.cl_partition(p3)))
##### control parameters
nclasses=9
nnruns=50
contr=list(verbose=TRUE,nruns=nnruns,k=nclasses)
meth="hard/euclidean"
##### fit consensus partition
pcons=cl_consensus(p,method=meth,control=contr)
print(pcons)
##### compute unweighted criterion measure
Crit=rep(0,length(p))
for (i in 1:length(p))
{
tmp=cl_dissimilarity(pcons,p[[i]],method="euclidean")
Crit[i]=tmp[[1]]
}
cat("Manually computed criterion: ")
cat(as.character(sum(Crit)))
______________________________________________
R-help@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.
