SAS has something called the "cubic criterion" cutoff for finding the most appropriate number of clusters. Does R have anything that would replicate that? I've been searching the lists and can't seem to find anything that would point me in the right direction. Thank in advance, Philip Bermingham
Le 17.05.2005 14:42, Philip Bermingham a ??crit :> SAS has something called the "cubic criterion" cutoff for finding the > most appropriate number of clusters. Does R have anything that would > replicate that? I've been searching the lists and can't seem to find > anything that would point me in the right direction. > > Thank in advance, > Philip Bermingham >Hello, Package fpc has a function cluster.stats with a lot of criterion like G2, G3, etc ... Romain -- ~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~ ~~~~~~ Romain FRANCOIS - http://addictedtor.free.fr ~~~~~~ ~~~~ Etudiant ISUP - CS3 - Industrie et Services ~~~~ ~~ http://www.isup.cicrp.jussieu.fr/ ~~ ~~~~ Stagiaire INRIA Futurs - Equipe SELECT ~~~~ ~~~~~~ http://www.inria.fr/recherche/equipes/select.fr.html ~~~~~~ ~~~~~~~~ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ~~~~~~~~
Hello, It depends, *which clustering method you will use*. Model-based Clustering algorithms have the BIC criterion implemented (Mclust). Partition Based clustering algorithms have other criterias (Sum of Squares withhin and between clusters and you can easely implement other criterias). Most of the criterias in fuzzy clustering are very different. With hierarchical clustering algorithms, you can also determine the number of cluster, very different from the other methods. Note that there is no optimal criteria for all these different methods and it is not so easy to find the optimal number of clusters - the optimal number of clusters depends on your data and which criteria you have to use depends also on your data. Best, Matthias> SAS has something called the "cubic criterion" cutoff for finding the > most appropriate number of clusters. Does R have anything that would > replicate that? I've been searching the lists and can't seem to find > anything that would point me in the right direction. > > Thank in advance, > Philip Bermingham > > ______________________________________________ > 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 >
Dear Philip, Perhaps this reference might be useful: V. D. P. Pillar. How sharp are classifications? Ecology 80:2508-2516, 1999. Adriano S. Melo>------------------------------ > >Message: 9 >Date: Tue, 17 May 2005 08:42:02 -0400 >From: Philip Bermingham <pberming at research.ryerson.ca> >Subject: [R] Finding the right number of clusters >To: r-help at stat.math.ethz.ch >Message-ID: <4289E69A.7040300 at csca.ryerson.ca> >Content-Type: text/plain; charset=ISO-8859-1; format=flowed > >SAS has something called the "cubic criterion" cutoff for finding the >most appropriate number of clusters. Does R have anything that would >replicate that? I've been searching the lists and can't seem to find >anything that would point me in the right direction. > >Thank in advance, >Philip Bermingham > >Adriano Sanches Melo Dep. Ecologia, Instituto de Bioci?ncias Universidade Federal do Rio Grande do Sul Porto Alegre, RS, 91540-000, BRAZIL http://www.ecologia.ufrgs.br/~adrimelo/e/index.htm -------------- next part -------------- No virus found in this outgoing message. Checked by AVG Anti-Virus.