Setting range for simultaneous histogram RooFit

Greetings:

I am trying to set a range for a simultaneous combined histogram fit in RooFit. There are two separate things that look like they should work based on the documentation I have read, but they aren’t working. I will try to explain this in steps for a simplified test I am running.

What does work is the simultaneous fit without a range:

I have defined pdfs using my own generated pdf (mypolypdf) and generated histograms according to quadratic, cubic, and quartic polynomials:

Note that “<” and “>” don’t appear when I post this for the RooFit Result and std::map lines below.

//Histograms:
 RooRealVar *myvar = new RooRealVar("myvar", "myvar",-range, range);
 RooDataHist *myroohistquad = new RooDataHist ("myroohistquad","myroohistquad",*myvar,Import(*myhistquad));
 RooDataHist *myroohistcubic = new RooDataHist ("myroohistcubic","myroohistcubic",*myvar,Import(*myhistcubic));
  RooDataHist *myroohistquartic = new RooDataHist ("myroohistquartic","myroohistquartic",*myvar,Import(*myhistquartic));

//Categories:
  RooCategory polytype("polytype","polytype");
  polytype.defineType("quadratic");
  polytype.defineType("cubic");
  polytype.defineType("quartic");

//Combined histogram
  std::map <std::string,RooDataHist*> histset;
  histset["quadratic"] = myroohistquad;
  histset["cubic"] = myroohistcubic;
  histset["quartic"] = myroohistquartic;
  RooDataHist CombPolyHist ("CombPolyHist", "CombPolyHist", *myvar, Index(polytype),Import(histset));

//pdfs:
  mypolypdf mypdfrooquad("mypdfrooquad","mypolypdf",*myvar,*p1roo,*p2roo,*p3rooquad,*p4rooquad);
  mypolypdf mypdfroocubic("mypdfroocubic","mypolypdf",*myvar,*p1roo,*p2roo,*p3roocubic,*p4roocubic);
  mypolypdf mypdfrooquartic("mypdfrooquartic","mypolypdf",*myvar,*p1roo,*p2roo,*p3rooquartic,*p4rooquartic);

  RooSimultaneous polysimulpdf("polysimulpdf", "polysimulpdf", polytype);
  polysimulpdf.addPdf(mypdfrooquad,"quadratic");
  polysimulpdf.addPdf(mypdfroocubic,"cubic");
  polysimulpdf.addPdf(mypdfrooquartic,"quartic");

//Fit:
  std::unique_ptr<RooFitResult> fitResult{polysimulpdf.fitTo(CombPolyHist, PrintLevel(-1), Save())};

This works fine.

Test 1: Make each histogram a function of a different variable:

Replaced histogram definitions above with:

  RooRealVar *myvarquad = new RooRealVar("myvarquad", "myvarquad",-range, range);
  RooRealVar *myvarcubic = new RooRealVar("myvarcubic", "myvarcubic",-range, range);
  RooRealVar *myvarquartic = new RooRealVar("myvarquartic", "myvarquartic",-range, range);
RooDataHist *myroohistquad = new RooDataHist ("myroohistquad","myroohistquad",*myvarquad,Import(*myhistquad));
RooDataHist *myroohistcubic = new RooDataHist ("myroohistcubic","myroohistcubic",*myvarcubic,Import(*myhistcubic));
RooDataHist *myroohistquartic = new RooDataHist ("myroohistquartic","myroohistquartic",*myvarquartic,Import(*myhistquartic));

  RooDataHist CombPolyHist ("CombPolyHist", "CombPolyHist", RooArgSet(*myvarcubic,*myvarquad,*myvarquartic), Index(polytype),Import(histset));

Trying this gives the following error:

[#0] ERROR:InputArguments -- Layout or binning of histogram myroohistquad is not consistent with first processed histogram myroohistcubic
Error in <TRint::HandleTermInput()>: std::invalid_argument caught: Layout or binning of inputs for RooDataHist is inconsistent

Test 2:
Keep the same variable for all histograms, and set the range on that variable:
Keep the original code but replace the fitTo with the following:

  myvar->setRange("highrange_quadratic",0.0,10.0);
  myvar->setRange("highrange_cubic",0.0,10.0);
  myvar->setRange("highrange_quartic",0.0,10.0);
  std::unique_ptr<RooFitResult> fitResult{polysimulpdf.fitTo(CombPolyHist, Range("highrange"), SplitRange(kTRUE), Save(), PrintLevel(-1))};

This results in very strange fits over a range of the full histogram for the last histogram (quartic) and with a strange function plotted also outside the histogram (values of x coordinate below that in the plot) for teh other two.

I am using ROOT 6.30/04

Do you have any suggestions on how to fix these?

Thanks very much!

Hi,

Thanks for the post.
My first suggestion is to use code formatting in the posts: it’s far easier to read reproducers and ultimately helps converging on the solution of the problem being discussed, which is in the interest of everybody :slight_smile:

Said that, I am adding in the loop our RooFit experts, @jonas and @moneta .

Cheers,
D