From: modlin1@metrocast.net   
      
   BlankThinking about a possible data mining technique in very high   
   dimensional data I worked around to a configuration that I now recognize as   
   approximately equivalent to doing PCA on overlapping subsets of the   
   variables, then doing PCA again on the combined outputs from the first   
   "layer" of PCA's.  (Possibly continuing through more layers if there are   
   truly a huge number of variables.)   
      
   This seems such a simple idea that it must have been done, and probably   
   exhaustively analyzed.   
      
   But so far I have not been able to find anything about this kind of   
   multi-stage PCA.  Probably just looking in the wrong places.   
      
   Could anyone suggest a source or provide an offhand analysis of the results   
   to be expected?   
      
   Without actually doing it and before I recognized the similarity to PCA I   
   thought it looked very promising.  But now that I see this way of describing   
   it, it seems unlikely to be a new approach.   
      
   Bill Modlin   
      
   [ comp.ai is moderated.  To submit, just post and be patient, or if ]   
   [ that fails mail your article to , and ]   
   [ ask your news administrator to fix the problems with your system. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)