Re: K-Means Clustering

[John Darrington]

Hi Mehmet,

Thanks for this.  It seems to be basically working.  There are a number of 
improvements
that can be made however.

1. It'll be more consistent with the rest of PSPP if you call the new file 
"quick-cluster.c"
   with a hyphen.

2. Instead of editing the Makefile, add the name of the new file to the 
manifest in 
   src/language/stats/automake.mk

3. Can you remove the line UNIMPL_CMD ("QUICK CLUSTER", "Fast clustering")
   from command.def instead of commenting it out.



 Now the "quick cluster" command can parse these options in the pspp command 
line:
     
     quick cluster /VARIABLES=x y z /GROUPS=5 /MAXITER=100.

This is different to the syntax in the SPSS documentation which expects:

   QUICK CLUSTER x y z 
      /CRITERIA = CLUSTER(5) MXITER (100).

where the /CRITERIA subcommand and each part thereof is optional.  You can see 
an example of how to 
implement a /CRITERIA subcommand in src/language/stats/factor.c - in fact, you 
may be able to copy much of that parser's code.

Avoid using atoi in the parser.  Instead of    groups=atoi(lex_tokcstr(lexer));
write :
        lex_force_int (lexer);
        groups = lex_integer (lexer); 
     

  i think a small development pdf documentation does not satisfies the needs of 
implementing 
  something in PSPP.

You're  right.  The developer documentation is woefully incomplete.

You mentioned earlier that you had tested the results against spss. Do you have 
the results
from these tests, and the test data that you used?  I would be interested to 
see this.
     

Best regards,

John

-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://pgp.mit.edu or any PGP keyserver for public key.

signature.asc
Description: Digital signature