leasqr and residuals

[dantebz]

Hi, I have a problem with the leasqr function, to be more specific with it's
residual calculation.
It returns a "weighted residuals are not real" error, at leasrq.m line 654
and subroutine __lm_svd__.m line 462.
I'm trying to use it to get a model for the function below:

model_function=@(x,p) -log10(1+power((x/p(2)),p(1))) +p(3);

where x is a Distance vector ranging from ~10Km to ~1500Km, and p is a
parâmeter, the result (data) is a vector of magnitude ranging from -6 to
-11, I've ran the program sucessfully for a specific frequency, while I
limited the the p(3) to lower values.

What I really want is to know how leasqr and __lm_svd__ (sub function)
calculate those residuals, and which values for 'p' are causing the unreal
residuals.

I'm a rookie user of Octave, and this is my first topic, so, sorry in
advance for any mistake or misuse of this tool.



--
View this message in context: 
http://octave.1599824.n4.nabble.com/leasqr-and-residuals-tp4667776.html
Sent from the Octave - General mailing list archive at Nabble.com.