I normally make a plot of the Wilk's ratio against the number
of classes. (Wilks=det(W)/det(T), W: pooled within-class
cov. matrix; T: total cov. matrix). This plots would be a negative
exponential for a totally unstructured tree, so the plot allows one
to identify singular points where an increase of 1 more
class implies a large decrease of the Wilk's ratio.
Agus
Dr. Agustin Lobo
Instituto de Ciencias de la Tierra (CSIC)
Lluis Sole Sabaris s/n
08028 Barcelona SPAIN
tel 34 3 330 2716
fax 34 3 411 0012
alobo@ieec.fcr.es
alobo@eno.princeton.edu
This archive was generated by hypermail 2b30 : Sat Feb 10 2001 - 10:21:31 MET