Hi Bob,
Thanks for reporting this problem. Now fixed in the CVS head.
Rene Brun
On Tue, 18 Sep 2001, Bob Schaefer wrote:
> Hello,
>
> I seemed to find 2 problems while using root to fit histograms. 1) Funny
> behavior in fitting sums of functions when a landau function is involved, and
> 2) Getting the SetOptFit to display the parameters as described.
>
>
> 1) I tried fitting a sum ("total") of a landau function and an exponential
>
> TH1F *hist;
>
> // fitting functions
> TF1 *exp1, *land, *total;
> // fitting parameters
> Double_t par[5];
>
> land = new TF1("land", "landau(0)",20,800);
> hist->Fit(land,"WQ0","",300,800);
> hist->Fit(exp1,"WQ0+","",20,200);
>
> exp1 = new TF1("exp1","expo(0)", 20, 800);
>
> land->GetParameters(&par[0]);
> exp1->GetParameters(&par[3]);
>
> total->SetParameters(&par[0]);
>
> total = new TF1("total","expo(3)+landau(0)", 20, 800);
> hist->Fit(total,"WR");
>
> This works fine unless I try to do the fitting with the exponential parameters
> FIRST in the parameter list, i.e.,
>
> land->GetParameters(&par[2]);
> exp1->GetParameters(&par[0]);
>
> total->SetParameters(&par[0]);
> total = new TF1("total","expo(0)+landau(2)", 20, 800);
> hist->Fit(total,"WR");
>
> In which case the fit fails, complaining that the ERROR MATRIX IS NOT POSITIVE
> DEFINITE.
>
>
> 2) Fit parameters not being displayed
>
> According to the documentation:
>
> gStyle->SetOptFit(0101);
>
> should print the names/values of the parameters AND the Chi-squared. It
> does not print the chi-squared, although it does print the parameters.
>
> However, if I use
>
> gStyle->SetOptFit(10101);
>
> it prints both chi sq and the parameters?!
>
> Thanks,
> Bob Schaefer
>
>
>
>
>
> Root 3.1/02 on LinuxPPC2000 running on a Mac G4 dual processor.
>
>
> ----------------------------------------------------------
> Robert K. Schaefer Email: bob@milkyway.gsfc.nasa.gov
> NASA/GSFC Code 664 Phone: 301-286-0878
> Greenbelt, MD 20771 FAX: 301-286-1629
> ----------------------------------------------------------
>
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:51:00 MET