Hello,
Way back in 2001, Stefan Anderhalden reported a problem using the fanny
function in S-PLUS 2000: no matter how many clusters were requested (by setting
the input parameter k), the algorithm returned a cluster membership matrix that
was consistent with the presence of just two clusters. Testing the fanny
function with the built-in ruspini dataset with k=4, however, produced results
with four clusters identical to those shown in the manual (Ch. 4 of S-PLUS 2000
Guide to Statistics, Vol 2). A big difference between my data and the ruspini
data is that I have 89 variables and 329 observations in my data whereas the
ruspini data only has two variables and 75 observations. Some experimentation
with running fanny on various subsets of my data suggests that the membership
coefficients computed by fanny are not what one might expect when the dataset
exceeds a certain size.
Has anyone had similar experience with fanny? Have there been any bug reports
for fanny? I would appreciate any information. Thanks,
Till
**********************************************
Till E. Stoeckenius, Senior Consultant
ENVIRON
101 Rowland Way, Suite 220
Novato, CA 94945
415-899-0709 (voice) 415-899-0707(fax)
tstoeckenius@environcorp.com
**********************************************
This message contains information that may be confidential, privileged or
otherwise protected by law from disclosure. It is intended for the exclusive
use of the Addressee(s). Unless you are the addressee or authorized agent of
the addressee, you may not review, copy, distribute or disclose to anyone the
message or any information contained within. If you have received this message
in error, please contact the sender by electronic reply to
email@environcorp.com and immediately delete all copies of the message.
|