RE: Map the chi2

Hi Rene,

you mean something like this?  

TH2D *scan01 = new TH2D("scan01", "Scan par[1] vs par[0]", 17, 0.235, 0.405, 36, 2.45, 6.05);

Double_t p0p1p2[3] = {0., 0., cdfCrossSectionFit->GetParameter(2)};

Double_t calcChi2 = 0;

for (Double_t p0 = 0.24; p0 <= 0.41; p0 += 0.01)

  {

    p0p1p2[0] = p0;

    for (Double_t p1 = 2.5; p1 <= 6.; p1 += 0.1)

      {

        p0p1p2[1] = p1;

        gMinuit->Eval(0, ???, calcChi2, p0p1p2, 3);

        scan01->SetBinContent(p0, p1, calcChi2);

      }

  }

scan01->Draw("SURF2");    

Where can I get the second argument for Eval?  

Bye...      

> -----Original Message-----

> From: Rene Brun [mailto:Rene.Brun_at_cern.ch]

> Sent: Tuesday, March 11, 2008 08:37

> To: Manuel Mussini

> Cc: 'Lorenzo Moneta'; roottalk_at_lxbuild091.cern.ch

> Subject: Re: [ROOT] Map the chi2

>

> Hi Manuel,

>

> What you require should be quite easy using:

> gMinuit->Eval(Int_t npar, Double_t *grad, Double_t &fval, Double_t *par,

> Int_t flag);

>

> Create a TH3 (or TGraph2D), make your nested 3 loops on each parameter

> in the limits you want.

> In the loop call gMinuit->Eval and store the returned value fval in your

> TH3 or TGraph2D.

>

> Rene Brun

>

> Manuel Mussini wrote:

> > Hi Lorenzo,

> > contours scans only "near" the found minimum so I think it isn't the right

> > choice for searching for another eventual "far" minimum!

> > I've never tried SCAN! I'll do it as my next step!

> > Is it possible to scan "by hand"? I mean ... can I fix all my 3 parameters

> and

> > evaluate the chi2? With TMinuit::FixParameter(Int_t) I think I can get the

> first

> > part but ... how to update the chi2?

> > The best result would be to obtain a 3D histo showing the chi2 surface (for

> each

> > parameters couple)!

> >

> >

> > Bye.

> >

> >

> >

> > ________________________________________

> > From: Lorenzo Moneta [mailto:Lorenzo.Moneta_at_cern.ch]

> > Sent: Monday, March 10, 2008 23:14

> > To: Manuel Mussini

> > Cc: roottalk_at_lxbuild091.cern.ch

> > Subject: Re: [ROOT] Map the chi2

> >

> > Hi Manuel,

> >

> > you can either scan a single parameter around the minimum, (command SCAN in

> > Minuit) or build contours for two parameters, like in fitcont.C.

> > I think contours will give you the best way for maping the chi2 around the

> > minimum.

> >

> > Cheers

> >

> > Lorenzo

> > On Mar 10, 2008, at 7:20 PM, Manuel Mussini wrote:

> >

> >

> > Hi,

> > I have a TH1D and a 3 parameters user defined TF1 to fit it.

> > Which is the best way to map the chi2 around the convergence point?

> > I thoughed to map the 3 parameters in pairs (fix one of them and scan the

> other

> > two) but it not so clear (to me!) how to implement the mapping!

> > I've tried the fitcont.C tutorial but this is not what I want!

> >

> > Thanks!!!

> >

> > Bye.

> >

> >

> >
  Received on Tue Mar 11 2008 - 09:39:38 CET