[SciPy-user] curve_fit failure when supplying uncertainties

[ElMickerino]

I think I've identified a problem with curve_fit which occurs when one
attempts to fit normally distributed data with a gaussian.  From the
documentation of curve_fit, it appears that 'sigma' should be the
uncertainties on the y-values of the data; however, the following example
(see attached code) should make it clear that there's a problem with this.  

My best guess is that the sigma are actually weights (=1.0/sigma).  Can
anyone confirm or deny this?  Seems like from the name it should be
uncertainties but from the behavior of the code it appears otherwise.

Also, I was wondering if there's a way to supply asymmetric errors to
curve_fit (or for that matter, to leastsqr or any wrapper thereof).

Thanks very much,
Michael

http://www.nabble.com/file/p22380378/testFit.py testFit.py 
-- 
View this message in context: http://www.nabble.com/curve_fit-failure-when-supplying-uncertainties-tp22380378p22380378.html
Sent from the Scipy-User mailing list archive at Nabble.com.