R: Can I 'set' my dendrogram with a for loop using dendextend?

Question

Given the iris data set as an example, I do Pearson clusters using dendextend in the following way:

library(RColorBrewer)
library(dendextend)
data(iris)
newmat <- iris[,1:4]
rownames(newmat) <- paste(iris$Species, rownames(iris))

dmat <- 1 - cor(t(newmat), method="pearson")
dmat <- as.dist(dmat)
clust.obj <- hclust(dmat, method="complete")
dend.obj <- as.dendrogram(clust.obj)
numsamples <- length(rownames(newmat))
maxdist <- max(get_nodes_attr(dend.obj, "height"))

groups <- levels(iris$Species)
cols <- colorRampPalette(brewer.pal(length(groups), "Set1"))
myPal <- cols(length(groups))

vals1 <- grep(groups[1], labels(dend.obj), value=TRUE)
vals2 <- grep(groups[2], labels(dend.obj), value=TRUE)
vals3 <- grep(groups[3], labels(dend.obj), value=TRUE)
vals1B <- grepl(groups[1], labels(dend.obj))
vals2B <- grepl(groups[2], labels(dend.obj))
vals3B <- grepl(groups[3], labels(dend.obj))

dend.obj <- dend.obj %>%
  set("leaves_pch", 19) %>%
  set("leaves_cex", 1) %>%
  set("branches_lty", 2) %>%
  set("by_labels_branches_col", value = vals1, TF_values = c(myPal[1],Inf)) %>%
  set("by_labels_branches_lwd", value = vals1, TF_values = c(2,Inf)) %>%
  set("by_labels_branches_lty", value = vals1, TF_values = c(1,Inf)) %>%
  set("by_labels_branches_col", value = vals2, TF_values = c(myPal[2],Inf)) %>%
  set("by_labels_branches_lwd", value = vals2, TF_values = c(2,Inf)) %>%
  set("by_labels_branches_lty", value = vals2, TF_values = c(1,Inf)) %>%
  set("by_labels_branches_col", value = vals3, TF_values = c(myPal[3],Inf)) %>%
  set("by_labels_branches_lwd", value = vals3, TF_values = c(2,Inf)) %>%
  set("by_labels_branches_lty", value = vals3, TF_values = c(1,Inf)) %>%
  set("labels_colors", ifelse(vals1B, myPal[1], ifelse(vals2B, myPal[2], myPal[3]))) %>%
  set("leaves_col", ifelse(vals1B, myPal[1], ifelse(vals2B, myPal[2], myPal[3])))

png(filename="test.png", height=1200, width=400)
mar.default <- c(5,4,4,2) + 0.1
par(mar = mar.default + c(0, 0, 0, 4))
plot(dend.obj, main="test cluster", xlab="Distance", horiz=TRUE, cex.main=1, cex.axis=1, cex.lab=1)
legend(maxdist, numsamples, groups, cex=1, pch=19, col=myPal)
dev.off()

Which produces this cluster, which I find really useful with the colors and all:

The problem is that now I want to encapsulate this into a function. And the length of groups can vary. So I need to do the set part of dend.obj inside a for loop or something.

Something like:

for (i in 1:length(groups)){
    set("by_labels_branches_col", value=grep(groups[i],labels(dend.obj),value=TRUE), TF_values=c(myPal[i],Inf)) %>%
    set("by_labels_branches_lwd", value=grep(groups[i],labels(dend.obj),value=TRUE), TF_values=c(2,Inf)) %>%
    set("by_labels_branches_lty", value=grep(groups[i],labels(dend.obj),value=TRUE), TF_values=c(1,Inf))
}

Which obviously does not work... And the same goes for the ifelse in there, which would be really tricky.

Any help would be appreciated! I have no idea how to tackle this. Thanks!

I'm not totally clear on what you are trying to do but would it be easier if you made another function that did the set part and then apply it to however many groups there are? It might be easier to understand if you showed what you have so far for the function — see24, Aug 29 '18 at 13:27
I am just trying to do the exact same, but without the knowledge that `groups` (`levels(iris$Species)`) has length 3... something that can be applied to `groups` of any length — DaniCee, Aug 30 '18 at 01:46
@Anony-Mousse try the code, it gives different errors and warnings, particularly `Error in plot.window(...) : need finite 'xlim' values` when trying to plot it — DaniCee, Aug 30 '18 at 01:50

score 0 · Accepted Answer · answered Aug 30 '18 at 03:29

0

Alright, in case anyone is interested, what I have done at the end, which seems to be working properly, is do the for loop and store the commands in an execval variable, then run that variable with:

eval(parse(text=execval))

answered Aug 30 '18 at 03:29

DaniCee

2,397
6
36
59

Glad you found a solution (although it feels like something that could probably be done without it. But you'll need another usecase to justify us going back to the question :) ) Cheers. – Tal Galili Aug 31 '18 at 16:27

R: Can I 'set' my dendrogram with a for loop using dendextend?

1 Answers1