Dear RooFit experts,
I´m performing a simultaneous fit between 10 histograms and I have a problem in constructing the data sample.
When I construct the sample I´m forced to use the ctor
RooDataHist(const char* name, const char* title, const RooArgList& vars, RooCategory& indexCat, mapstd::string,RooDataHist* dhistMap, Double_t wgt = 1.0)
instead of
RooDataHist(const char* name, const char* title, const RooArgList& vars, const RooCmdArg& arg1, const RooCmdArg& arg2 = RooCmdArg(), const RooCmdArg& arg3 = RooCmdArg(), const RooCmdArg& arg4 = RooCmdArg(), const RooCmdArg& arg5 = RooCmdArg(), const RooCmdArg& arg6 = RooCmdArg(), const RooCmdArg& arg7 = RooCmdArg(), const RooCmdArg& arg8 = RooCmdArg())
because I have more tha 8 arguments (I have 10 distributions). I perform the fit with the code below.
RooRealVar numTrack("numTracks","#Tracks",0,20);
//The fitted variable
RooCategory IDcat("IDcat","IDcat");
IDcat.defineType("NoID");
IDcat.defineType("IDCutL");
IDcat.defineType("IDCutM");
IDcat.defineType("IDCutT");
IDcat.defineType("IDLlhL");
IDcat.defineType("IDLlhM");
IDcat.defineType("IDLlhT");
IDcat.defineType("IDbdtL");
IDcat.defineType("IDbdtM");
IDcat.defineType("IDbdtT");
RooSimultaneous simPDF("simPDF","simPDF",IDcat);
simPDF.addPdf(sumTemNoID,"NoID");
simPDF.addPdf(sumTemIDCutL,"IDCutL");
simPDF.addPdf(sumTemIDCutM,"IDCutM");
simPDF.addPdf(sumTemIDCutT,"IDCutT");
simPDF.addPdf(sumTemIDLlhL,"IDLlhL");
simPDF.addPdf(sumTemIDLlhM,"IDLlhM");
simPDF.addPdf(sumTemIDLlhT,"IDLlhT");
simPDF.addPdf(sumTemIDbdtL,"IDbdtL");
simPDF.addPdf(sumTemIDbdtM,"IDbdtM");
simPDF.addPdf(sumTemIDbdtT,"IDbdtT");
// sumTemX is the template I use for the model in each cathegory
map<string,TH1*> hmap ;
hmap.insert(std::pair<string, TH1*>("NoID", h_numTrackNoID));
hmap.insert(std::pair<string, TH1*>("IDCutL", h_numTrackIDCutL));
hmap.insert(std::pair<string, TH1*>("IDCutM", h_numTrackIDCutM));
hmap.insert(std::pair<string, TH1*>("IDCutT", h_numTrackIDCutT));
hmap.insert(std::pair<string, TH1*>("IDLlhL", h_numTrackIDLlhL));
hmap.insert(std::pair<string, TH1*>("IDLlhM", h_numTrackIDLlhM));
hmap.insert(std::pair<string, TH1*>("IDLlhT", h_numTrackIDLlhT));
hmap.insert(std::pair<string, TH1*>("IDbdtL", h_numTrackIDbdtL));
hmap.insert(std::pair<string, TH1*>("IDbdtM", h_numTrackIDbdtM));
hmap.insert(std::pair<string, TH1*>("IDbdtT", h_numTrackIDbdtT));
// h_numTrackX is the data TH1F for each category
RooDataHist simData("simData","simData",numTrack,IDcat,hmap,1./10);
simPDF.fitTo(simData,Minos(kTRUE));
You may note that in the RooDataHist ctor I put the weight 1/10. Otherwise the errors estimated by the fit are largely underestimated and the statistics in each plot I draw showing the various fitted distributions for each category, seems to be multiplied by 10. I think that using this weight is a safe way to do the fit, because, trying to use for example only 4 distributions, and building the RooDataHist with the second ctor (being able to in this case) the results are the same of using the first ctor putting the weight 1/4.
What is the reason for this weight I must put in?
Can I trust the results of my fit?
Thank You very much,
Federico