[eigen] [Eigen] NonLinearOptimization unsupported module: How to know to goodness of fit?

[ Thread Index | Date Index | More lists.tuxfamily.org/eigen Archives ]

Dear all,

I am using the NonLinearOptimization module to perform some non-linear fittings on a data set (a simulated pdf distribution).
I am fitting both the theoretical pdf and the ccdf (ccdf = 1 - cdf, cdf = Cummulative distribution function). But, I am getting different
values for the distribution parameters. And the parameters error after the fit are small for both cases.
I would like to know if there is a criteria I can use to know the goodness of the fit. After checking the test cases for the NonLinearOptimzation
module, I noted that the blueNorm of the fvec member of the LevenbergMarquardt object was always compared with some small values,
and actually for my case it gives me some clues (the ccdf fit produces a smaller value for that quantity than the pdf fit), but I am unsure if this
 can be taken as the criteria or if there is another one already implemented on the non linear fitting routine.

Thanks in advance for your help.

Best regards / Cordialmente,

William-Fernando Oquendo
Este correo puede carecer de tildes o eñes debido al teclado. 

Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/