Dear R-Users and Developers, I want to calculate something like k-means clustering, but with ordinal data (Braun-Blanquet) to combine this classification technique with a NMDS-Ordination. I found an algorithm especially developed for categorical data: k-modes (Huang 1998). Is there a function implemented in R which can calculate with this algorithm. I found a hint on a poster-abstract from Gero Szepannek, Uwe Ligges and Claus Weihs. There it is stated that the algorithm is implemented in their package klaR. But I cannot find anything about k-modes in the klaR-Tutorial. Could you please help me. Thanks Kim -- GRATIS f?r alle GMX-Mitglieder: Die maxdome Movie-FLAT!
Kim Vanselow wrote:> Dear R-Users and Developers, > I want to calculate something like k-means clustering, but with ordinal data (Braun-Blanquet) to combine this classification technique with a NMDS-Ordination. > I found an algorithm especially developed for categorical data: k-modes (Huang 1998). Is there a function implemented in R which can calculate with this algorithm. I found a hint on a poster-abstract from Gero Szepannek, Uwe Ligges and Claus Weihs. There it is stated that the algorithm is implemented in their package klaR. But I cannot find anything about k-modes in the klaR-Tutorial. > Could you please help me.Dear Kim, well, maybe you start writing a message to the authors of that poster-abstract at first who might be able to help? Since I am one of those: Where can I find this abstract or poster (at least in order to remind Gero who does no longer work at the university to finally implement the feature, if it is not yet there)? Best wishes, Uwe Ligges> > Thanks > Kim
Uwe Ligges wrote:> > > Kim Vanselow wrote: >> Dear R-Users and Developers, >> I want to calculate something like k-means clustering, but with >> ordinal data (Braun-Blanquet) to combine this classification technique >> with a NMDS-Ordination. >> I found an algorithm especially developed for categorical data: >> k-modes (Huang 1998). Is there a function implemented in R which can >> calculate with this algorithm. I found a hint on a poster-abstract >> from Gero Szepannek, Uwe Ligges and Claus Weihs. There it is stated >> that the algorithm is implemented in their package klaR. But I cannot >> find anything about k-modes in the klaR-Tutorial. >> Could you please help me. > > Dear Kim, > > well, maybe you start writing a message to the authors of that > poster-abstract at first who might be able to help? > Since I am one of those: Where can I find this abstract or poster (at > least in order to remind Gero who does no longer work at the university > to finally implement the feature, if it is not yet there)?Some news on this: Gero never presented it in his talk at the useR!2008 and we all forgot that he announced it in his abstract. He is working for a bank these days and is going to send his code within a few days. Will take a look if it is easy to add the existing code to klaR then. Best wishes, Uwe Ligges> Best wishes, > Uwe Ligges > > >> >> Thanks >> Kim >
Uwe Ligges wrote:> > > Uwe Ligges wrote: >> >> >> Kim Vanselow wrote: >>> Dear R-Users and Developers, >>> I want to calculate something like k-means clustering, but with >>> ordinal data (Braun-Blanquet) to combine this classification >>> technique with a NMDS-Ordination. >>> I found an algorithm especially developed for categorical data: >>> k-modes (Huang 1998). Is there a function implemented in R which can >>> calculate with this algorithm. I found a hint on a poster-abstract >>> from Gero Szepannek, Uwe Ligges and Claus Weihs. There it is stated >>> that the algorithm is implemented in their package klaR. But I cannot >>> find anything about k-modes in the klaR-Tutorial. >>> Could you please help me. >> >> Dear Kim, >> >> well, maybe you start writing a message to the authors of that >> poster-abstract at first who might be able to help? >> Since I am one of those: Where can I find this abstract or poster (at >> least in order to remind Gero who does no longer work at the >> university to finally implement the feature, if it is not yet there)? > > Some news on this: > > Gero never presented it in his talk at the useR!2008 and we all forgot > that he announced it in his abstract. He is working for a bank these > days and is going to send his code within a few days. Will take a look > if it is easy to add the existing code to klaR then.FYI: I got the code now, will try to incorporate it soon. Best, Uwe> Best wishes, > Uwe Ligges > > >> Best wishes, >> Uwe Ligges >> >> >>> >>> Thanks >>> Kim >> >