Re: Clarification about fit errors

Hi Alberto,

  the procedure used to fit a TGraphError is explained in http://root.cern.ch/root/htmldoc/TGraph.html#TGraph :Fit
(see the paragraph "How errors are used in the chisquare function")

  The resulting parameter errors should not really depend on the parameter range. Only if your parameter is at the range limit then the error is not (and cannot be ) correctly calculated. If you see an effect for parameter values not at the limit, please send me a script reproducing this effect so I can investigate it

  Best Regards

  Lorenzo
On Oct 10, 2009, at 12:32 PM, Alberto Pulvirenti wrote:

> Dear all,
>
> firs of all I would like to understand the standar procedure which
> is carried on when I use the "fit" method on TGraphError objects.
> Can someone help me in this?
>
> Moreover, I would like to understand the reason of the fit behaviour
> when I define a range for some parameters. What happens in this case
> is that the errors on these ranged parameters, which I set using
> TF1::SetParLimits, depend on the range I set: the smaller the range,
> the smaller the error. I would like to understand then how these
> errors are determined and how I can obtain some kind of "absolute"
> errors even with ranged parameters.
>
> Thanks a lot and best regards
>
> Alberto
>
Received on Sun Oct 11 2009 - 22:48:37 CEST