confusion matrix analysis

[James Darrell McCauley]

Attn: remote sensing types

would there be any interest for a grass module to do iterative
proportional fitting with smoothing of zeros with pseudo-counts?
m.ipf? This is a program to normalize error or confusion matrices
(contingency tables) resulting from an image classification.

BTW, is there an alternative to r.coin to produce a contingency table
comparing two raster maps? Functionally, r.coin does exactly what I want
but it adds a lot of "extraneous formatting." I'm looking for
something to produce a simple matrix like

  a b c
a 3 1 2
b 0 4 0 
c 1 1 4

where [a-c] are category numbers and [0-4] are coincidence values (or
maybe even leave off the category numbers).  Perhaps someone has added
an extra flag to do this (?). Is there another grass command that
can do this easily? I ask because, if I produce m.ipf, I 
would like to generate the input matrix easily.

--Darrell

P.S. on a similar note, has anyone written an m.kappa to compute
kappa (and do hypothesis testing)?