[ROOT] SetObjFit / TMinuit

From: Antonio Kanaan (kanaan@astro.ufsc.br)
Date: Wed Jun 20 2001 - 12:59:38 MEST


Hi All,

I have been using minuit to fit multiple Gaussians.  My code is
becoming more elegant now and I don't want to keep using global
variables so I dug in TMinuit.cxx and found this at the beggining:

//*-*   - The functions SetObjFit(TObject *obj) and GetObjFit() can be    *
//*-*     used inside the FCN function to set/get a referenced object     *
//*-*     instead of using global variables.                              *

I have no clue as what to do to actually make use of this.  I found
the definition (?) of this function in G__Minuit.cxx but that only
made me more confuse.  I guess first of all the my object containing
the data to be fit needs to be transformed to a TObject and then I can
use SetObjectFit (note there is no SetObjFit - I am saying this not to
be picky just to point out a little inconsistency between the comment
and the code) within FCN to have access to my object.  However, how do
I get this object inside FCN?  I must confess there is a couple
arguments in the FCN function whose meaning I never quite understood:

void fcn(Int_t &npar, Double_t *gin, Double_t &f, Double_t *par, Int_t iflag)
{
  // computes Chi^2
  // &npar = number os parameters 
  // *gin  = ???????? , not used here
  // &f = returns the chisq^2 value in this variable
  // *par = the array of parameters
  // iflag = ???????? , not used here

so far I have survived without messing with them...  But neither one
is of type TObject, therefore I gather they are not what I am after.  

If anyone has some example code showing the use of SetObjectFit
(hopefully showing the creation of the TObject as well) I would
appreciate it.  

thanks, have you all a nice day,

Antonio


Antônio Kanaan
Departamento de Física - Universidade Federal de Santa Catarina
CP 476 --  CEP 88040-900  --  Florianópolis  --  SC  --  Brasil
http://www.astro.ufsc.br/~kanaan   e-mail: kanaan@astro.ufsc.br
Phone: 55,48,3319069                       FAX  : 55,48,3319946



This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:50:49 MET