Re: [ROOT] Fit TGraphError

From: Rene Brun (Rene.Brun@cern.ch)
Date: Thu Feb 15 2001 - 22:21:42 MET


Hi Stephanie,
Note that the option "B" does (did) not exist in TGraph::Fit.
When fitting with predefined functions like poln,expo,gaus,
the inititial values of the parameters are automatically computed.
However, some time it may be useful to switch off this automatic
computation. I have added an option "B" in TH1::Fit and
TGraph::Fit to disable this automatic initial values.
The new option is now in CVS.

Rene Brun

On Thu, 15 Feb 2001, S. Escoffier wrote:

> Dear Rooters,
> 
> I would like to report a bug in fitting a TGraphError:
> If I fit a TGraphError with a polynomial function, with the parameter 0 fixed
> to 0. ,
> then results are not in agreement with initials values. I mean, p0 is well
> fixed, but not equal to 0.
> If function f1 is described by
>     TF1 *f1 = new TF1("f1","[0]+[1]*x",0.,5.);
> then it works.
> 
> I'm working with 2.23.12 version on UNIX system.
> Sorry if this problem has been already reported.
> 
> Stephanie.
> 
> 
> void test()
> {
>     Float_t x[4]={1.,2.5,3.5,4.75};
>     Float_t ex[4]={0.01,0.01,0.01,0.01};
>     Float_t y[4]={0.,1.,2.,3.};
>     Float_t ey[4]={0.01,0.01,0.01,0.01};
> 
>    TGraphErrors* g1 =new TGraphErrors(4,x,y,ex,ey);
> 
>     TF1 *f1 = new TF1("f1","pol1",0.,5.);
>     f1->SetParameter(0,0.);
>     f1->SetParLimits(0,1,1);
> 
>     TCanvas* c1;
>     c1 = new TCanvas("c1","c1",100,10,800,500);
>     c1->cd();gStyle->SetOptFit(111);
>     g1->Draw("AP");
>     g1->Fit("f1","B");
> 
> }
> 
> Print on screen:
> 
> FCN=2442.08 FROM MIGRAD    STATUS=CONVERGED      18 CALLS          19 TOTAL
>                      EDM=4.66971e-08    STRATEGY= 1      ERROR MATRIX ACCURATE
> 
>   EXT PARAMETER                                   STEP         FIRST
>   NO.   NAME      VALUE            ERROR          SIZE      DERIVATIVE
>    1  p0          -2.75862e-01     fixed
>    2  p1           6.47669e-01   1.72543e-03   4.16384e-05  -1.77117e-01
> 
> 



This archive was generated by hypermail 2b29 : Fri Jun 08 2001 - 11:51:09 MEST