different fit behaviour in ROOT and ROOFIT

From: Roberta Arnaldi <arnaldi_at_to.infn.it>
Date: Wed, 27 Jul 2011 17:28:43 +0200


Dear All,

I'm fitting an invariant mass spectrum histogram with a sum of functions (Crystal Ball for the peak and two exponentials for the background), using ROOT or ROOTFIT.
Initial parameters, fit limits...are the same in both cases.

The value I want to extract is the number of signal events under the Crystal Ball (CB) and the width of the Crystal Ball itself. Even if the fit quality is good in both cases, the number of events under the Crystal Ball is systematically different (~6-10% depending on the fit), as well as the CB width. In particular, ROOFIT gives always lower values. As an example:

Do you have any suggestion? I'm using root v5.28.00d and in ROOT I use the options "LI".
I attach to the mail the root file and two macro I use to perform the fit with root or roofit.

   Best regards and thanks in advance for your help,

             Roberta

Received on Wed Jul 27 2011 - 17:24:27 CEST

This archive was generated by hypermail 2.2.0 : Wed Jul 27 2011 - 17:50:01 CEST