R devel - Dec 2009 - Flexclust barchart issue when mcol=NULL (PR#14150)

Full_Name: Chris Hane
Version: 2.10.1
OS: Windows
Submission from: (NULL) (198.203.181.181)


When using barchart in the flexcust package, setting mcol=NULL to avoid the
lollipops causes an error.  Each panel shows the text message "Error using
packet n replacement has length zero." where n is the panel number.
> data(iris)
> cl <- cclust(iris[,-5], k=3)
> barplot(cl, mcol=NULL)  #works fine
> barchart(cl, mcol=NULL) # oops
FWIW, as the data of the clusters is scaled showing the colMeans for the centers
is misleading.  The global mean of each column is 0, not the colMean of the
centers.

Please do read the section on 'BUGS' in the FAQ.
This is not the bug-reporting address for package 'flexclust'.

On Wed, 16 Dec 2009, christopher.a.hane at gmail.com wrote:
> Full_Name: Chris Hane
> Version: 2.10.1
> OS: Windows
> Submission from: (NULL) (198.203.181.181)
>
>
> When using barchart in the flexcust package, setting mcol=NULL to avoid the
> lollipops causes an error.  Each panel shows the text message "Error
using
> packet n replacement has length zero." where n is the panel number.
>
>> data(iris)
>> cl <- cclust(iris[,-5], k=3)
>> barplot(cl, mcol=NULL)  #works fine
>> barchart(cl, mcol=NULL) # oops
>
> FWIW, as the data of the clusters is scaled showing the colMeans for the
centers
> is misleading.  The global mean of each column is 0, not the colMean of the
> centers.
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

>>>>> On Wed, 16 Dec 2009 07:05:11 +0100 (CET),
>>>>> christopher a hane (cah) wrote:
  > Full_Name: Chris Hane
  > Version: 2.10.1
  > OS: Windows
  > Submission from: (NULL) (198.203.181.181)


  > When using barchart in the flexcust package, setting mcol=NULL to avoid
the
  > lollipops causes an error.  Each panel shows the text message "Error
using
  > packet n replacement has length zero." where n is the panel number.

  >> data(iris)
  >> cl <- cclust(iris[,-5], k=3)
  >> barplot(cl, mcol=NULL)  #works fine
  >> barchart(cl, mcol=NULL) # oops

Yes, that's a bug in flexclust. I'll fix it for the next release.

In the meantime you can use barchart(cl, mcol=NA) which has the same
effect for me. 
  

  > FWIW, as the data of the clusters is scaled showing the colMeans
  > for the centers is misleading.  The global mean of each column is
  > 0, not the colMean of the centers.

Hmm, I am not sure I understand this paragraph. I agree that showing
the global mean makes no sense for centered data, but the iris data
are not centered. I don't think that barcharts like the above make
sense in all possible situations. I mainly use them for binary data
where I want to see which variables have a high/low percentage of ones
in which cluster (=bars), and what the difference to the global mean
is.

Best,
Fritz