>From: Madeleine Seeland <madeleine.seeland <at> in.tum.de>
>Subject: help with hclust
>Date: 2011-09-14 08:14:48 GMT
>Hello,
>
>I have two questions regarding hclust:
>
>1) First I would like to cut a hclust tree at a specific height and
receive the cluster (di) similarity at this
>height.
With an example:
hc <- hclust(dist(USArrests), "ave")
plot(hc, hang = -1)
hc.cf <- cophenetic(hc) #cophenetic distance
# cut cluster at height=50
rect.hclust(hc, h = 50)
hc.50 <- cutree(hc, h=50)
hc.simi <- NULL
for(i in unique(hc.50)) hc.simi <- c(hc.simi,
max(as.matrix(hc.cf)[names(hc.50[hc.50==i]),names(hc.50[hc.50==i])]))
hc.simi
[1] 44.28392 44.83793 20.59851 38.52791 41.09477
>2) Next, I would like to receive all clusterings up to a given tree
height that fulfill a minimal required
>similarity (i.e., the inner cluster similarity of all clusters
needs to fulfill a given similarity
>value). How can I achieve that?
For example, all clusterings with a (di)similarity at least 40
# which clusters have a maximum dissimilarity < 40
hc.le.40 <-unique(hc.50)[hc.simi<40]
# clusterings up to a height of 40
hc.50[hc.50==hc.le.40[1]|hc.50==hc.le.40[2]]
Connecticut Florida Idaho Indiana Kansas
3 4 3 3 3
Kentucky Montana Nebraska North Carolina Ohio
3 3 3 4 3
Pennsylvania Utah
3 3
HTH,
MArcelino
>
>I would appreciate your help!
>
>Best regards,
>Madeleine
>
____________________________________
Marcelino de la Cruz Rot
Depto. Biologia Vegetal
EUIT Agricola
Universidad Politecnica de Madrid
tel: 34 + 913365435
____________________________________
[[alternative HTML version deleted]]