Re: [ROOT] Fitting Option and Errors.

From: Rene Brun (Rene.Brun@cern.ch)
Date: Thu Oct 07 2004 - 09:24:15 MEST


Hi Zaldy,

zaldy wrote:
> 
> Dear folks,
> 
> I am just wondering why the result (particularly the fitted
> parameter errors) of
> 
> (a) TH1::Fit("myFunc") and
> (b) H1::Fit("myFunc","I")
> 
> seems to have little difference (~1% or less).
> 
> I expect a larger difference (not at this 1% level) since (b) accounts the
> bin errors (???)

I think that you misunderstood the option "I". When this option is specified,
the integral of the function for each bin is used instead of computing
just the function in the center of the bin. This option is more precise, but
also more time consuming as it implies computing many times the integral of the
function.
If you specify the option "I" or not, the bin errors are taken into account.
If your histogram has no associated errors (TH1::Sumw2 has not been called),
the errors are assumed to be the sqrt(bin content), otherwise the error
is the sqrt(sum of squares of weights per bin).

When using the option "E", ROOT will perform a better parameter error
estimation.
For more details, see the documentation of TMinuit, section error estimation.

Rene Brun

> 
> My questions:
> 1. In (a), what or how is the bin errors used in the fitting?
>    Is this the default (sqrt of bin content?)
> 
> 2. In (b), how do we qoute the fitted parameters if we change the option
> to "E"? I noticed that the MINOS errors(NEGATIVE/POSITIVE) are not exactly
> same (at least slightly differ).
> 
> Like to understand these points. Any Hints?
> 
> Thanks in Advance...
> 
> --
> Zaldy A. Nawang



This archive was generated by hypermail 2b29 : Sun Jan 02 2005 - 05:50:09 MET