[R] possible bug in flexclust

[Friedrich.Leisch at R-project.org]

>>>>> On Wed, 4 Jun 2008 13:40:58 -0400,
>>>>> Gary Lewis (GL) wrote:

  > Hi - Writing to see if someone can suggest whether a problem warrants a bug
  > report. It concerns the use of stepFlexclust in the flexclust package. The
  > problem concerns the size of clusters returned.

  > Versions: R-2.7.0 on Windows XP; RODBC_1.2-3

  > code snippet:
  > r8 <- stepFlexclust(df,8,nrep=100,FUN=kcca, family=kccaFamily("kmedians"))
  > summary(r8)  ## returns cluster sizes of 51, 115, 218, 73, 118, 140, 311,
  > and 118
  > trev <- data.frame(r8 at cluster)
  > sqlSave(ch, trev, rownames="unitid")

  > in MySQL:
  > select r8cluster, count(*) from trev group by 1 order by 1   ## returns
  > cluster sizes of 51, 116, 217, 73, 117, 140, 312, 118

  > problem:
  > 4 of the 8 clusters differ in sizes. They're close but I'd think they should
  > be exact.


Strange. What numbers do you get from

	 table(r8 at cluster)

? Which version of flexclust?

Best,
Fritz

-- 
-----------------------------------------------------------------------
Prof. Dr. Friedrich Leisch 

Institut für Statistik                          Tel: (+49 89) 2180 3165
Ludwig-Maximilians-Universität                  Fax: (+49 89) 2180 5308
Ludwigstraße 33
D-80539 München                     http://www.statistik.lmu.de/~leisch
-----------------------------------------------------------------------
   Journal Computational Statistics --- http://www.springer.com/180 
          Münchner R Kurse --- http://www.statistik.lmu.de/R