R help - Feb 2010 - How to: Compare Two dendrograms (Hierarchical Clusterings) ?

Hello all,

I wish to compare two dendrograms (representing Hierarchical Clusterings).

My problems are several:

1) how do I manually create a dendrogram object ?
That is, how can I reconstruct it as an "hclust" object that creates
such a
dendrogram, when all I have is the dendrogram image (but don't have the
underlaying distance matrix that produced it) ?
I see that there is a function called "as.dendrogram" But I wasn't
able to
find an example on how to use it.

2) Assuming that I have to hclust (or dendrogram ) objects, how can I
compare them ?
After some searching and asking, I found this article:
 http://www.jstor.org/pss/2288117
And a few others that cite it, that gives ideas on how to do this.
But I wasn't able to come a cross a R implementation of it.
Is there ?


Many thanks,
Tal



----------------Contact
Details:-------------------------------------------------------
Contact me: Tal.Galili@gmail.com |  972-52-7275845
Read me: www.talgalili.com (Hebrew) | www.biostatistics.co.il (Hebrew) |
www.r-statistics.com (English)
----------------------------------------------------------------------------------------------

	[[alternative HTML version deleted]]

Hi,

To compare two dendrogamas, you can use a measure of correlation between the
cophenetic matrices. Example:

mdist <- dist(iris[,-5], method="euclidean")
# dendrogram from cluster 1 (single-linkage)
hc1 <- hclust(mdist, method="single")
plot(hc1)

# dendrogram from cluster 2 (complete-linkage)
hc2 <- hclust(mdist, method="complete")
plot(hc2)

# correlation
cor(cophenetic(hc1),cophenetic(hc2))

For a confidence level, use the "Mantel Test" from package vegan.

mantel(cophenetic(hc1), cophenetic(hc2))

Good Look.