On the one hand, a number of variables
in X may not be informative about the clustering as present in the data, or, in
the worst case, may even mask the underlying clustering; as a consequence,
performing first a clustering in such a case may yield an incorrect clustering
(Vichi and Kiers 2001).