[statistics] computing properties on repeated data ?

[CdeMills]

Hello,
I have input-output measurements which I would like to characterise by a
covariance matrix. The trick is that each entry in the input sequence is
repeated, so I can compute a raw variance estimate of the output sequence,
for each level of the input sequence. From those estimates I can compute a
variance matrix for the whole output sequence. Such a function would have
two inputs, a data sequence and a grouping sequence. The target is to deal
with data which are not homoscedastic. Does such function already exists? If
not, should they belong to base-octave or octave-forge ?

Regards

Pascal
-- 
View this message in context: 
http://octave.1599824.n4.nabble.com/statistics-computing-properties-on-repeated-data-tp3092379p3092379.html
Sent from the Octave - General mailing list archive at Nabble.com.