Re: [ROOT] Errors on fit functions

From: Rene Brun (Rene.Brun@cern.ch)
Date: Thu Jul 24 2003 - 08:30:22 MEST


Hi Jason,

There is no such function in ROOT returning the fit error for a given 
value of x.
You have to use the cov matrix and compute the function value
with ffgr->Eval(x).

If you implement this function, send it to me. I will include it in the 
source.

Rene Brun

On Sun, 20 
Jul 2003, Jason Seely wrote:

> hi ROOT people,
> 
> i am just wondering if there is an easy way to get the error of a fit. 
>  i'm fitting something like this (for example)
> 
>     ffgr = new TGraphErrors(imax, x, y, dx, dy);
>     ffgr->Fit(func,"","",xmin,xmax);
> 
>     TF1    *fitfunc;
>     fitfunc = ffgr->GetFunction(func);
> 
> so once i have the fitting function, i know how to get parameters, 
> chi^2, parameter errors, etc.  what i want is the estimated error on the 
> function itself so i can draw an error band.  i've seen a lot of stuff 
> in ROOTtalk about how to get the covariance matrix, so i can do that to 
> calculate the error myself from the covariance matrix, but it seems like 
> there must be something like
> 
>     ffgr->GetError(x), or ffgr->EvalError(x), or whatever
> 
> to get the error as calculated from the covariance matrix so i don't 
> have to do all the work myself.  any ideas?
> 
> thanks,
> 
> jason
> 



This archive was generated by hypermail 2b29 : Thu Jan 01 2004 - 17:50:13 MET