Computing uncertainties of coefficients from fitting functions

[tstanley]

Does anyone know of a way to get the uncertainties in the coefficients from the fitting functions in the Advanced Analysis Library?  For example, for the LinFit function I'm looking for the uncertainties of both the slope and the intercept.  For the linear fit, I could just write my own, but I'm also looking for the uncertainties of the coefficients from the PolyFit and the NonLinearFit functions which would be more challenging.  If there was some kind of library I could use that did that it would save me a lot of time.

Thanks in advance!

[Wolfgang]

Hi,

There is no possibility using the NI routines, to my knowledge. The only thing you may do is calculate R^2 using the routine GoodnessOfFit.

For the individual error estimates one would need the covariance matrix, which is not availabe in the routines offered by NI.

My way out was to buy the Intel Math Library which provides many nice and efficient routines; unfortunately, linking these libraries is not simple (i.e. I failed)

Wolfgang

[tstanley]

Thanks for your comments.  Right now I'm looking at calling NIST's DataPlot application from my CVI application.  Not every elegant, but it should work.  Nice part about DataPlot is that it is free to download, use, and redistribute.  I'll also look into the Intel Math Library, but if you couldn't get it linked in CVI I kind of doubt that I could.  

[SoundNLight]

Here's a LabVIEW implementation of the linear least squares fitting routine from Bevingon & Robinson, 3rd edition, chapter 6, that yields the uncertainties in the fit coefficients.