Show dendrogram node values in R

Question

I am using the hclust function:

points <- data.frame(ID = c('A','B','C','D','E'), 
                 x = c(3,4,2.1,4,7), 
                 y = c(6.1,2,5,6,3))
d <- dist(as.matrix(points[, 2:3])) 
clusters <- hclust(d,method = "complete")
plot(clusters, labels=points$ID)

Is there a way to show the values where the points are joined (or the node values (where the dissimilarity between the samples is minimal))?

I want my plot to look like the one on the picture.

Note: The values showed on the dendrogram are not the correct ones.

gsun · Accepted Answer · 2017-04-26 20:33:29Z

My R package TBEST has a function that can add two color annotations to a hclust object. For your convenience, I am pasting codes below, so you can use them independent of any packages.

hc2axes<-function (x) {
    A <- x$merge
    n <- nrow(A) + 1
    x.axis <- c()
    y.axis <- x$height
    x.tmp <- rep(0, 2)
    zz <- match(1:length(x$order), x$order)
    for (i in 1:(n - 1)) {
       ai <- A[i, 1]
       if (ai < 0) 
           x.tmp[1] <- zz[-ai]
       else x.tmp[1] <- x.axis[ai]
           ai <- A[i, 2]
       if (ai < 0) 
           x.tmp[2] <- zz[-ai]
       else x.tmp[2] <- x.axis[ai]
           x.axis[i] <- mean(x.tmp)
    }
    return(data.frame(x.axis = x.axis, y.axis = y.axis))
}



plot_height<-function (hc, height, col = c(2, 3), print.num = TRUE, float = 0.01, cex = NULL, font = NULL) 
{
    axes <- hc2axes(hc)
    usr <- par()$usr
    wid <- usr[4] - usr[3]
    bp <- as.character(round(height,2))
    rn <- as.character(1:length(height))
    bp[length(bp)] <- "height"
    rn[length(rn)] <- "edge #"
    a <- text(x = axes[, 1], y = axes[, 2] + float * wid, bp, 
        col = col[1], pos = 2, offset = 0.3, cex = cex, font = font)
    if (print.num) {
        a <- text(x = axes[, 1], y = axes[, 2], rn, col = col[2], 
            pos = 4, offset = 0.3, cex = cex, font = font)
    }
}

Once you paste these two functions, add one line to plot your dendrogram, plot(clusters,labels=points$ID);
cluster_height(clusters,height=clusters$height,print.num=F)

You can also plot the branch numbers by setting print.num=T

MrFlick · Accepted Answer · 2017-04-26 19:41:56Z

4

Here's one method using the dendextend package.

First, convert to hanging dendrogram

library(dendextend)
dend <- as.dendrogram(clusters) %>% hang.dendrogram()
dend <- dend %>% set_labels(points$ID[dend %>% labels()])

Now we find the x,y values for all the internal nodes

xy <- dend %>% get_nodes_xy()
is_internal_node <- is.na(dend %>% get_nodes_attr("leaf"))
is_internal_node[which.max(xy[,2])] <- FALSE
xy <- xy[is_internal_node,]

And now we plot the dendrogram and draw the labels at a slight offset

plot(dend)
text(xy[,1]+.2, xy[,2]+.2, labels=format(xy[,2], digits=2), col="red")

This gives the following plot

edited Apr 26, 2017 at 19:41

answered Apr 26, 2017 at 19:36

MrFlick

209k19 gold badges300 silver badges324 bronze badges

2 Comments

Jane Over a year ago

Thank you, just what I needed! This is the first time I hear about this library and it seems awesome.

MrFlick Over a year ago

If you mean the labels we drew with text(), then set the cex= parameter. If you mean the leaf labels, check out the dendextend introduction, it's covered there.

Collectives™ on Stack Overflow

Show dendrogram node values in R

2 Answers 2

Comments

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related