plot.hclust: strange behaviour with "manufactured"

Thu, Feb 21, 2002 5:13 AM

This worked for me with your example:

source("dumpdata.R")
storage.mode(x.hc$merge) <- "integer"
plot(x.hc)

(R-1.4.1 compiled from source on WinNT4.)

Andy

-----Original Message-----
From: Hugh Chipman [mailto:hachipma at icarus.math.uwaterloo.ca]
Sent: Wednesday, February 20, 2002 5:32 PM
To: andy_liaw at merck.com
Cc: r-help at stat.math.ethz.ch
Subject: Re: [R] plot.hclust: strange behaviour with "manufactured"

Andy:

That's a great idea.  And it's certainly part of the problem.  I
verified this by changing the type of the "merge" element of a valid
hclust object to be double.  As you suggest, it doesn't plot.  

What's interesting is that if I change it back to integer again,
plot.hclust still gives an error.  Moreover, the object.size of the
"merge" component that was transformed integer to double to integer is
different from the object.size of the original integer component.  Is
there something that I am missing about integer types?

The original integer type was produced by the interal (C or Fortran)
code, so maybe I am corrupting it in some way.  

An example below gives the details...
--------------------------------------------------------------
--------

library(mva)
set.seed(103)
x <- matrix(rnorm(5 * 2), ncol = 2)
x.hc <- hclust(dist(x), method = "average")
plot.hclust(x.hc)
unlist(lapply(x.hc, storage.mode))

      merge      height       order      labels      method   
     call
  "integer"    "double"   "integer"      "NULL" "character"  
"language"
dist.method
"character"

unlist(lapply(x.hc, object.size))

      merge      height       order      labels      method   
     call
        152          60          52           0          44   
      324
dist.method
         52

x.hc$merge <- as.double(x.hc$merge)
plot.hclust(x.hc)

Error in plot.hclust(x.hc) : invalid dendrogram

unlist(lapply(x.hc,storage.mode))

 
      merge      height       order      labels      method   
     call
   "double"    "double"   "integer"      "NULL" "character"  
"language"
dist.method
"character"

unlist(lapply(x.hc,object.size))

 
      merge      height       order      labels      method   
     call
         92          60          52           0          44   
      324
dist.method
         52

x.hc$merge <- as.integer(x.hc$merge)
plot.hclust(x.hc)

 
Error in plot.hclust(x.hc) : invalid dendrogram

unlist(lapply(x.hc,storage.mode))

 
      merge      height       order      labels      method   
     call
  "integer"    "double"   "integer"      "NULL" "character"  
"language"
dist.method
"character"

unlist(lapply(x.hc,object.size))

 
      merge      height       order      labels      method   
     call
         60          60          52           0          44   
      324
dist.method
         52

The problem seems to be that the original merge component

is stored as int,

but the source()'d version is not.  Coercing that to int I

think works.

Cheers,
Andy

-----Original Message-----
From: Hugh Chipman [mailto:hachipma at icarus.math.uwaterloo.ca]
Sent: Wednesday, February 20, 2002 1:32 PM
To: r-help at stat.math.ethz.ch
Cc: hachipma at icarus.math.uwaterloo.ca
Subject: [R] plot.hclust: strange behaviour with

"manufactured" hclust

object


I've been trying to get plot.hclust to work with a hclust object I
created and have not had much success.  It seems that

there is some

"hidden" characteristic of a hclust object that I can't

see.  This is

most easily seen in the following example, where

plot.hclust works on

one object, but when this object is "dumped" and then re-read,
plot.hclust no longer works.  Is this a scoping problem?

Script started on Wed Feb 20 13:17:37 2002
1>@icarus[101]% R

R : Copyright 2002, The R Development Core Team
Version 1.4.1  (2002-01-30)

......(stuff deleted)

set.seed(103)
library(mva)
x <- matrix(rnorm(5*2),ncol=2)
x.hc <- hclust(dist(x),method='average')     
plot.hclust(x.hc)

# the above works

dump('x.hc')
source('dumpdata.R')
plot.hclust(x.hc)

Error in plot.hclust(x.hc) : invalid dendrogram input

One hypothesis I had was that it might be something like

the "call"

object that wasn't correctly being re-created by the dump

& source,

but that doesn't seem to be it.  Below, I remove everything except
merge, order, and height, and plot.hclust still works on

the original

object and fails on the dump + source object.

x.hc <- hclust(dist(x),method='average')
x.hc$call <- NULL
x.hc$dist.method <- NULL
x.hc$method <- NULL
x.hc$labels <- NULL
plot.hclust(x.hc)
dump('x.hc')
source('dumpdata.R')
plot.hclust(x.hc)

Error in plot.hclust(x.hc) : invalid dendrogram input

str(x.hc)

List of 3
 $ merge : num [1:4, 1:2] -2 -5 -1 -4 -3 1 2 3
 $ height: num [1:4] 0.873 1.257 1.633 2.430
 $ order : num [1:5] 4 1 5 2 3
 - attr(*, "class")= chr "hclust"

Any help would be greatly appreciated!

Hugh Chipman

-- 
-- Hugh Chipman -- Associate Professor, Statistics & 
Actuarial Science --
-- University of Waterloo, Ontario, Canada                    
         --
--                                                            
         --
-- Visiting Stanford Statistics Department, January-March 
2002         --
-- (650) 725 6148                                   Fax (650) 
725-8977 --
-- Department of Statistics, Sequoia Hall, 390 Serra Mall     
         --
-- Stanford University, Stanford, CA 94305-4065               
         --
-- http://www.stats.uwaterloo.ca/~hachipma/     
hachipman at uwaterloo.ca --
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.
-.-.-.-.-.-.-.-.-
r-help mailing list -- Read 
http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: 
r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._.
_._._._._._._._._


-- 
-- Hugh Chipman -- Associate Professor, Statistics & 
Actuarial Science --
-- University of Waterloo, Ontario, Canada                    
         --
--                                                            
         --
-- Visiting Stanford Statistics Department, January-March 
2002         --
-- (650) 725 6148                                   Fax (650) 
725-8977 --
-- Department of Statistics, Sequoia Hall, 390 Serra Mall     
         --
-- Stanford University, Stanford, CA 94305-4065               
         --
-- http://www.stats.uwaterloo.ca/~hachipma/     
hachipman at uwaterloo.ca --

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._