Cutting a dendrogram in R

Question

I am trying to cut this dendrogram into 3 groups: (T24, T1, T17, etc.), (T12, T15, T6, etc.) and (T2, T8, T3, T9)

I have tried using cutree(hc, k=3, h=400) , but it continues to make the same groups. Any help is greatly appreciated. Here is my code.

#temps must have date/time as column headers, not row headers
load(temps)
distMatrix <- dist(temps)
#create label colors
labelColors = c("#E41A1C", "#377EB8", "#4DAF4A", "#984EA3", "#FF7F00", "#FFFF33")
# cut dendrogram in 3 clusters
clusMember = cutree(hc, k=3, h=400)
colLab <- function(n) {
  if (is.leaf(n)) {
  a <- attributes(n)
  labCol <- labelColors[clusMember[which(names(clusMember) == a$label)]]
  attr(n, "nodePar") <- c(a$nodePar, lab.col = labCol)
  }
  n
}
hcd = as.dendrogram(hc)
clusDendro = dendrapply(hcd, colLab)
plot(clusDendro, main = "Cluster Analysis")

You can*t, because there is not height (y axis) that represents those three groups. — lukeA, Jan 21 '16 at 21:16
You can just do a smaller height, and combine groups yourself afterwards. The docs for cuttree say if you specify `k` it overrides `h` if both are given. — Andy W, Jan 21 '16 at 21:23

score 4 · Answer 1 · answered Jan 22 '16 at 08:50

You example is not reproducible in the sense that we don't have access to the data. What I can say is that you should look at the dendextend R package. It offers functions for cutting a dendrogram, as well as coloring labels and branches. The Quick Introduction manual shows the basic use of functions such as labels_colors and color_branches for producing plots such as this:

In your case, since your branches seem to be in the same height, you are not likely to have direct control over their cuts. What you could do is use the branches_attr_by_clusters to specifically control the the colors of the sub-clusters you care for. Here is an example:

x <- c(1:3, 6:8)
dend <- as.dendrogram(hclust(dist(x), method = "ave"))
library(dendextend)
labels(dend) <- x[order.dendrogram(dend)]

# due to the ties - there is specific reason to have this be these 3 clusters:
cutree(dend, k = 3)[order.dendrogram(dend)]

par(mfrow = c(1,2))
dend1 <- color_branches(dend, k = 3)
dend1 <- color_labels(dend1, k = 3)
plot(dend1, main = "default cuts by cutree")
# let's force it to be another 3 clusters:
dend2 <- branches_attr_by_clusters(dend, c(1, 2,2, 3,3,3), c("gold", "darkgreen", "blue"))
# coloring the labels is actually the easiest part:
labels_colors(dend2) <- c("gold", "darkgreen", "blue")[c(1, 2,2, 3,3,3)]
plot(dend2, main = "Manual cuts")

how to use color_branches after this to add for heatmap.2 `ColSideColors`? — Brian Wiley, Dec 21 '20 at 04:26
Nevermind it would be `labels_colors(dend2)[order(order.dendrogram(dend2))]` — Brian Wiley, Dec 21 '20 at 04:28
I think the labels need to be unique here also in `labels(dend)` — Brian Wiley, Dec 21 '20 at 05:57
Confirmed unique or doesn't have to be unique but with the correct order of the manual breaks. — Brian Wiley, Dec 21 '20 at 06:06

Cutting a dendrogram in R

1 Answers1