I'm using COA to compare the result of the classification of a
multi-temporal satellite image to
a set of 7 landcover maps of the Mediterranean Basin.
I have produced 7 contingency matrices and run COA on
each one.
I have a couple of questions.
1. Can I use the value of the first eigenvalue as
a measure of correpondence? Note that although
the contingency matrix has always the same number of rows
(because I'm always using the same classification), the number
of columns is different for each COA as the different landcover
maps have different legends with different number of
categories. Is there any way to take the number of columns
into account at evaluating the first eigenvalue?
I would say that I could use the ratio between the first observed
eigenvalue to the average eigenvalue from the permutation
test, would this be correct?
Or should I just use a Chi sq. coefficient?
2. If I plot the positions of rows and columns in the COA space
(i.e., the plane defined by the first two axes),
can the euclidean distances in the
COA space between a given row and a given column
be used as a measure of linkage? I think that there are some
problems with this practice.
Thanks,
Agus
Dr. Agustin Lobo
Instituto de Ciencias de la Tierra (CSIC)
Lluis Sole Sabaris s/n
08028 Barcelona SPAIN
tel 34 93409 5410
fax 34 93411 0012
alobo@ija.csic.es
http://pangea.ija.csic.es/alobo
This archive was generated by hypermail 2b30 : Sat Feb 10 2001 - 10:36:01 MET