R help - Aug 2010 - Grouping clusters from dendrograms

Hi,

I have produced a dendrogram of categorical data in R using the hclust
function, although the input was a dissimilarity matrix produced in SAS, as
I have defined my own distances. 

The dendrogram is fine and I can view and use this. However, I was wondering
if there is a method by which I can find out the optimal place to place
groups, rather than relying on my visual analysis? I don't have multiple
trees to compare and therefore would be relying on the data in the one tree. 

I have tried using dynamic tree cut, but this seems to be producing odd
results whereby it is grouping data that overlap branches. However, I'm not
sure if this is because of the script I am using to visualise the data which
may be incorrect. 

The script I tried was:
> dendro<-hclust(as.dist(combo2),method="average")
>
cut2<-cutreeDynamicTree(dendro,maxTreeHeight=1,deepSplit=TRUE,minModuleSize=1)
I then used the following to visualise the data:
> cut2colour<-labels2colors(cut2)
> plotDendroAndColors(dendro,cut2colour,"Dynamic Tree Cut",
> dendroLabels=FALSE,hang=0.03,addGuide=TRUE,guideHang=0.05)
Any advice or ideas would be much appreciated!

Thanks,

Julia
-- 
View this message in context:
http://r.789695.n4.nabble.com/Grouping-clusters-from-dendrograms-tp2316521p2316521.html
Sent from the R help mailing list archive at Nabble.com.

Hi Julia, 

sorry for the very late reply, your original email was posted while I was on
hiatus from R-help. I'm the author of the dynamicTreeCut package. I
recommend that you try using the "hybrid" method using the
cutreeDynamic
function. What you observed is a known problem of the tree method (which, by
the way, was the reason I developed the Hybrid method). 

Using the hybrid method is simple, for example as 

cut2<-cutreeDynamic(dendro,distM = combo2,
maxTreeHeight=1,deepSplit=2,minModuleSize=1)

You can play with the argument deepSplit to obtain finer or coarser modules.

HTH,

Peter 

--
View this message in context:
http://r.789695.n4.nabble.com/Grouping-clusters-from-dendrograms-tp2316521p3988526.html
Sent from the R help mailing list archive at Nabble.com.