RooFit categories and RooSimultaneous

ROOT
1

Hello,

I have a root-tree that I use to create a RooDataSet.

The branches I use, let’s say, are called mass and likelihood. I want to do a simultaneous mass fit to two sub-samples of the RooDataSet: one with likelihood > cut and the other with likelihood < cut.

I understand that I can make an intermediate step, re-create a Tree with additional branch that would be a discrete 0 or 1, and read it as a RooCatergory when I create the RooDataSet.

But is there an easier way of doing it? There are many advantages to doing this in one program…

Thanks,
Yuri

2

Hi Yuri,

You don’t have to recreate a tree, you just need to add a column to your
dataset that stored the decision of your cut in a RooCategory. Given

RooRealVar* mass, likelihood // variables in dataset, assumed to be declared
RooDataSet d // data with observables ‘mass’ and ‘likelihood’, defined earlier

you do something like this

// Define a real->category mapping splitting sample in “sampleA” and "sampleB"
RooThresholdCategory thresh_func(“tresh”,“thresh”,*likelihood,“SampleA”) ;
thresh_func.addThreshold(cutValue,“SampleB”) ;

// Add category to your data, return value is RooCategory with same states
// defined as input function
RooCategory* thresh = (RooCategory*) data->addColumn(thresh_func) ;

Then proceed to build a RooSimultaneous p.d.f with index category ‘thresh’

Wouter

3

[quote=“Wouter Verkerke”]
// Add category to your data, return value is RooCategory with same states
// defined as input function
RooCategory* thresh = (RooCategory*) data->addColumn(thresh_func) ;

Then proceed to build a RooSimultaneous p.d.f with index category ‘thresh’

Wouter[/quote]

but that does not work - that was one of the first things I tried. Problem is, Roo Simultaneous asks for RooAbsCategoryLValue&, and RooThresholdCategory does not inherit from it, unlike RooCategory.

Yuri

P.S. Any ideas about my other post? Subject “reading dataset from a tree”

4

Hi Yuri,

You should give RooSimultaneous the ‘tresh’ object return by addColumn()
rather than the thres_func object.

Wouter

5

Wouter,

thanks so much, the fit now works very well. Problem is, I’m now having trouble plotting the result. I want to plot m3 distributions passing and failing the cut with fit results overlaid. I get the data to plot correctly, but for the fit result I only get the “fail” component. Snippet of my code is below

RooRealVar efficiency(“efficiency”,“efficiency”,0.9,-1.0,1.0);
RooRealVar nSig(“nSig”,“nSig”,1000.0,-10.0,1000000.0);
RooFormulaVar nSigpass(“nSigpass”,“nSigefficiency", RooArgList(nSig,efficiency) ) ;
RooFormulaVar nSigfail(“nSigfail”,"nSig(1.0 - efficiency)”, RooArgList(nSig,efficiency) ) ;

RooArgList componentspass(sigShapePdf,bkgShapePdf);
RooArgList yieldspass(nSigpass, nBkgpass);
RooAddPdf sumpass(“sumpass”,“fixed extended sum pdf”, componentspass, yieldspass);

RooArgList componentsfail(sigShapeFPdf,bkgShapePdf );
RooArgList yieldsfail(nSigfail, nBkgfail );
RooAddPdf sumfail(“sumfail”,“fixed extended sum pdf”, componentsfail, yieldsfail);

RooThresholdCategory pf_cut("pf_cut","pf_cut",p_r9,"pass",0);
pf_cut.addThreshold(0.9,"fail",1);

RooCategory *MyfitCat = (RooCategory*) data->addColumn(pf_cut);
RooSimultaneous totalPdf("totalPdf","totalPdf",*MyfitCat);
totalPdf.addPdf(sumpass,"pass");
totalPdf.addPdf(sumfail,"fail");

    TCanvas *cc = new TCanvas();
    cc->Divide(1,2);
    cc->cd(1);
    RooPlot* frame1 = m3.frame(60);
frame1->SetTitle("");
data->plotOn(frame1,Cut("pf_cut==0"));
MyfitCat->setLabel("pass");
totalPdf.plotOn(frame1,Slice(*MyfitCat),ProjWData(m3,*data)) ;
    totalPdf.plotOn(frame1,Slice(*MyfitCat),Components(bkgShapePdf),LineStyle(kDashed),ProjWData(m3,*data)) ;
frame1->Draw("e0");

    cc->cd(2);
    RooPlot* frame2 = m3.frame(60);
frame2->SetTitle("");
data->plotOn(frame2,Cut("pf_cut==1"));
MyfitCat->setLabel("fail");
totalPdf.plotOn(frame2,Slice(*MyfitCat),ProjWData(m3,*data)) ;
    totalPdf.plotOn(frame2,Slice(*MyfitCat),Components(bkgShapePdf),LineStyle(kDashed),ProjWData(m3,*data)) ;
frame2->Draw("e0");
1 Like
6

Hi Yury,

Sorry for the late reply, I was offline last week.

I see nothing obviously wrong with your macro but I cannot reproduce
your problem as the macro below is incomplete (i.e. I miss data and definition of m3).
If you send me a self-contained example I will have look.

Wouter