[R] x/y coordinates of dendrogram branches

[Romain Francois]

Hi Timo,

Here is a quick and dirty recursive function.
##################
absi <- function(hc, level=length(hc$height), init=TRUE){
  if(init){
    .count <<- 0
    .topAbsis <<- NULL
    .heights <<- NULL
  }
  if(level<0) {
    .count <<- .count + 1
    return(.count)
  }
  node <- hc$merge[level,]
  le <- absi(hc, node[1], init=FALSE)
  ri <- absi(hc, node[2], init=FALSE)
 
  mid <- (le+ri)/2
 
  .topAbsis <<-  c(.topAbsis, mid)
  .heights <<- c(.heights, hc$height[level])
 
  invisible(mid)
}
#################

R> mydata <- c(1,2,3,4,5)
R> hc <- hclust(dist(mydata), method="average")
R> absi(hc)
R> .topAbsis # those are x-coordinates
[1] 1.500 4.500 3.750 2.625
R> .heights  # and y-coordinates
[1] 1.0 1.0 1.5 2.5
R> plot(hc)
R> points(.topAbsis, .heights, pch=21, bg=2, cex=5)


Is this close to what tou need ?


Romain




Le 02.11.2005 14:51, Timo Becker a écrit :

>Dear R-users,
>
>I need some help concerning the plotting of dendrograms for hierarchical 
>agglomerative clustering.
>The agglomeration niveau of each step should be displayed at the 
>branches of the dendrogram.
>For this I need the x/y coordinates of the branch-agglomerations of the 
>dendrogram.
>The y-values are known (the heights of the agglomeration), but how can I 
>get the x-values?
>
> > mydata <- c(1,2,3,4,5)
> > hc <- hclust(dist(mydata), method="average")
> > hc$height # these are the y-coordinates
>[1] 1.0 1.0 1.5 2.5
>
>I experimented with the dendrogram object because it gives the midpoints:
>
> > dend <- as.dendrogram(hc)
> > attributes(dend[[1]])$midpoint
>[1] 0.5
>
>Perhaps I could loop over all possible branches and get the midpoints if 
>I knew how to convert them to x-values.
>The manual says that midpoint is the "numeric horizontal distance of the 
>node from the left border (the leftmost leaf) of the
>branch (unit 1 between all leaves)."
>But how do I know which one is the leftmost leaf and what is its 
>horizontal value?
>
>Thanks in advance,
>Timo
>
>______________________________________________
>R-help at stat.math.ethz.ch mailing list
>https://stat.ethz.ch/mailman/listinfo/r-help
>PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
>
>
>  
>


-- 
visit the R Graph Gallery : http://addictedtor.free.fr/graphiques
+---------------------------------------------------------------+
| Romain FRANCOIS - http://francoisromain.free.fr               |
| Doctorant INRIA Futurs / EDF                                  |
+---------------------------------------------------------------+