hclust order and cutree groups in R

Asked Aug 31 '20 at 14:08

Active Mar 03 '21 at 15:55

Viewed 741 times

I want to create a data frame that has my row names, their order according to hclust, and their group membership according to cutree. I'm having a hard time reconciling the outputs from these functions.

d <- dist(USArrests)
hc <- hclust(d)
pheatmap::pheatmap(as.matrix(d)[hc$order, hc$order], cluster_cols = F, cluster_rows = F) # what I expect to see

data.frame(
  state = hc$labels,
  hc = cutree(hc, h = 0), # hc doesn't match heat map
  family = cutree(hc, k = 5) # family order doesn't agree with hc order
)

Based on my understanding of hierarchical clustering, hc and family should both be in the same order. For example:

hc     = 1 2 3 4 5 6 # smallest to greatest
family = 1 1 2 2 2 3 # smallest to greatest

Update:

This gets me the result I'm looking for, I want to see if there's a simpler way to do this

data.frame(state = hc$labels[hc$order],
  row = 1:length(hc$labels),
  family = cutree(hc, k = 5)[hc$order]
)

edited Sep 01 '20 at 15:59

asked Aug 31 '20 at 14:08

Jeff Bezos

1,929
13
23

Maybe `hc$labels[hc$order]` – GKi Aug 31 '20 at 14:12
That seems to match the heat map, now how do you get the `family` column to match up? – Jeff Bezos Aug 31 '20 at 14:14
1

Maybe `cutree(hc, k = 5)[hc$order]` – GKi Aug 31 '20 at 14:14
So `tibble(state = hc$labels[hc$order], order = 1:length(hc$labels), family = cutree(hc, k = 5)[hc$order][state])` – Jeff Bezos Aug 31 '20 at 14:18
@GKi your solution seems to work, I wonder if there's an easier way to get this result – Jeff Bezos Aug 31 '20 at 14:18
Maybe `data.frame(state = hc$labels, row = seq_along(hc$labels), family = cutree(hc, k = 5))[hc$order,]` – GKi Aug 31 '20 at 14:34

hclust order and cutree groups in R

0 Answers0