Hi Luigi, If you want to fit a function with many parameters to an histogram 1 or 2-d, I STRONGLY suggest to use TH1::Fit instead of implementing yourself the fitting procedure directly with Minuit. We have many examples with fitting in the tutorials. TH1::Fit (see also TGraph::Fit) implements a Chisquare method and also a Loglikelihood method for fitting. In case you would like to use TMinuit directly, you can find examples of use in the TH1 fit procedures. Look first at: http://root.cern.ch/root/htmldoc/src/TH1.cxx.html#TH1:Fit then look at the functions H1FitChisquare and H1FitLikelihood. Rene Brun Bardelli Luigi wrote: > > I'm trying a multidimensional fit using the TMinuit class with my custom > FCN to be minimized: after some documentation, these are the questions > left: > > - What is the exact meaning of the void MyFunc(Int_t &npar, Double_t *gin, > Double_t &f, Double_t *u, Int_t flag) arguments of the custom > function (set by SetFCN) ? > Moverover which is the input and which the output of MyFunc ? > > - Defining a parameter: the TMinuit::DefineParameter method allows to set > a parameter with limits; how can I set one parameter without? > > Asking for an advice: > Having a function with many parameters, and no analytical derivative, > which is the best approach to fit the data? Is it correct to use the > "raw" TMinuit ? > > (I'm using ROOT Version 3.00/02) > > Thanks for any help > > Luigi > > ------------------------------- > Luigi Bardelli > INFN Sezione di Firenze > E-Mail: bardelli@fi.infn.it > -------------------------------
This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:51:06 MET