RE: Using Parameters in TTreeFormula's?

From: Philippe Canal <pcanal_at_fnal.gov>
Date: Thu, 12 Oct 2006 16:41:42 -0500

Hi Robert,

> I do NOT want to 'fit' to the population's distribution like the
> example does -- the function I'm trying to use (TF1 or whatever)
> should NOT return something proportional to the number of events
> in a bin/region. Instead, the evaluation is similar to TGraph,
> where the Chi2 function is built directly from the X and Y
> points.

As far as I know/understand this is __exactly__ (except for actually creating the intermediary TGraph2D object) what TTree::UnbinnedFit does ....

Seriously, it sound that you want you exactly want is:

TF2 *myf = new TF2("myf","[0]*x+[1]*y",0.,3000.,0.,3000.); myf->SetParameters(.1,.1);
tree->UnbinnedFit("myf","electronP:preshowerE:showerE",selection);

Cheers,
Philippe.

-----Original Message-----
From: Robert Feuerbach [mailto:feuerbac_at_jlab.org] Sent: Thursday, October 12, 2006 2:01 PM To: Rene Brun
Cc: pcanal_at_fnal.gov; roottalk_at_pcroot.cern.ch Subject: Re: [ROOT] Using Parameters in TTreeFormula's?

Hi Rene,

Just to restate my original question about using parameters ('[0]', '[1]', etc.) like TFormula has in a TTreeFormula. Are there any plans to make them work in a TTreeFormula?

Thank you for the example but it is not applicable. I do NOT want to 'fit' to the population's distribution like the provided example does -- the function I'm trying to use (TF1 or whatever) does NOT return something proportional to the number of events in a bin/region but instead a value to be compared with one of the TTree variables.

This is a task that is performed more for calibration purposes, and not for extracting cross-sections and peaks. This application requires an evaluation similar to what is in TGraph, where the Chi2 function is built directly from the X and Y points.

For my current limited case with two 'independent' and one 'dependent' tree variable, the easiest solution I have found is to use a TGraph2D:

Int_t ne = T->Draw("electronP:preshowerE:showerE",selection);

Double_t *z = T->GetV1();
Double_t *y = T->GetV2();
Double_t *x = T->GetV3();

TGraph2D gra2(nv,x,y,z);

TF2 *myf = new TF2("myf","[0]*x+[1]*y",0.,3000.,0.,3000.); myf->SetParameters(.1,.1);
gr2a.Fit("myf2");

Still, a general solution would be nice to find and use, especially once there are more than two 'independent' variables.

Thank you,
Rob Feuerbach

On Thu, 12 Oct 2006, Rene Brun wrote:

> Date: Thu, 12 Oct 2006 06:00:34 +0200
> From: Rene Brun <Rene.Brun_at_cern.ch>
> To: Robert Feuerbach <feuerbac_at_jlab.org>
> Cc: Philippe Canal <pcanal_at_fnal.gov>, roottalk_at_pcroot.cern.ch
> Subject: Re: [ROOT] Using Parameters in TTreeFormula's?
>
> Robert,
>
> As already explained, you have everything you want in TTree::UnbinnedFit.
> In the doc of this function, we give an example with one variable
> // TF1* f1 = new TF1("f1", "gaus(0)/sqrt(2*3.14159)/[2]", 0, 5);
> // f1->SetParameters(1, 3.1, 0.01);
> // f1->SetParLimits(0, 1, 1); // fix the normalization parameter to 1
> // data->UnbinnedFit("f1", "jpsimass", "jpsipt>3.0");
>
> in case of 3 variables, give a TF3 as input using either an inline
> formula as above for the 1-d case
> or a Double_t myfunc(Double_t*x, Double_t*par).
> and call UnbinnedFit with something like
> data->UnbinnedFit(myfunc,"var1:var2:var3","some condition")
> where var1, var2, var3 are the 3 ntuple variables that you want to fit.
>
> Rene Brun
> Robert Feuerbach wrote:
> > Hi Philippe,
> >
> > Thank you for your indulgence, but from what I have seen in the
> > examples and my understanding of using TF1, what UnbinnedFit does
> > is fit a 'distribution' of events, not the relation between
> > different variables in the tree. If I am wrong, please point me
> > to an example of fitting the relation between two or more tree
> > variables.
> >
> > For now, for my specialized case, if TTreeFormula's Parameters
> > worked similarly to TFormula, I'm following the minexam.cxx
> > example and using my own fcnk0( ) which has to be modified by
> > hand for every case.
> >
> > As for an implementation of the 'Minimize' method/routine, it
> > would be something like what is below.
> >
> > Take care,
> > Rob Feuerbach

/***************************************************
 * Dr. Robert J. Feuerbach       feuerbac_at_jlab.org *
 * 12000 Jefferson Avenue       CEBAF Center  A120 *
 * Suite 4                  Office: (757) 269-7254 *
 * Newport News, VA 23606     Page:       584-7254 *
 * Fax: (757) 269-5703               Mailstop 12H3 *
 ***************************************************/

Received on Thu Oct 12 2006 - 23:42:14 MEST