comparing estimates

[CdeMills]

Hello,

I ask for help about statistical tests. 

The estimates are obtained as follows: I have a source signal S and a
measured channel M. The source signal contains 1000 samples; for which a
record 1000 values. Then I model
M = cst + linear_gain*S
I extract the cst and the gain by least squares, then I compute residuals as
M predicted - M Measured; error propagation technique finally leads to the
covariance matrix on the two estimates (cst; gain); their degrees of freedom
is 1000-2.

This experiment is repeated three times, and I would like to verify that the
differences between the three values of linear gain are compatibles with the
standard deviations obtaiend previously. It seems that ANOVA is the way to
go: I have to compare three estimates, but they already carry their own
variances  and degrees of freedom. How to compute the ANOVA table in this
case ?

Regards

Pascal

--
View this message in context: 
http://octave.1599824.n4.nabble.com/comparing-estimates-tp3902129p3902129.html
Sent from the Octave - General mailing list archive at Nabble.com.