forum.alglib.net

Hi,

I'm using Alglib for the non-linear fitting (C++).
Now I need to compute a confidence intarvals.
In order to do that I need to know the Standard Errors.
My question is which one from the report parameters should I use?
I'm talking about "lsfitfit" + "lsfitresults" functions.


I have already tried the "Rep.ErrCurve", however the CI seems too small in this case. From the documentation it seem like this is not the SE.
When using "Rep.Noise" it looks much better. Is that the one corresponding to per-point asymptotic standard deviations?
Or what does it mean "vector of per-point estimates of noise"? Could you clarify this a little?

--
Kind Regards,
Dmitry