Using RooRealVar in Integer steps

[tintin2807]

Dear Rooters,

I’m trying to generate a binomial pdf which I can then fit to my measured data. Problem is RooFit doesn’t seem to be able to compute the pdf in discrete integer steps, or to put it another way it has no functionality to allow the user to declare an integer variable, with RooRealVar() implicitly computing the pdf in arbitrary non-integer steps.

Using RooRealVar("","",RooConst(x),RooConst(x1),RooConst(x2)) gives error message …"candidate constructor not viable…

Using RooConstVar() gives similar message.

Is there a workaround? I posted a similar question on the “Stat and Math Tool Support” a few days ago but have had no takers ! perhaps I might have better joy here !

Many many thanks in advance!

---

see example of code and output below

#ifndef __CINT__
#include "RooGlobalFunc.h"
#endif
#include "RooRealVar.h"
#include "RooDataSet.h"
#include "TMath.h"
#include "TCanvas.h"
#include "RooPlot.h"
#include "TAxis.h"
using namespace RooFit ;

void FitGeneratedBinomialModelToData()
{
  // Declare variables
  RooRealVar p("p","p",0.5,0,1) ;
  RooRealVar smalln("smalln","smalln",1,0,50) ;
  RooRealVar bigN("bigN","bigN",40,39,41) ;
  // build pdf
 RooGenericPdf binomialpdf("binomial","binomial PDF","pow(p,smalln)*pow((1-p),(bigN-smalln))*TMath::Factorial(bigN)/(TMath::Factorial(smalln)*TMath::Factorial(bigN-smalln))",RooArgSet(smalln,p,bigN)) ;

  // Construct plot frame in 'm'
  RooPlot* xframe = smalln.frame(Title("Generated binomial p.d.f.")) ;
  // Plot
  binomialpdf.plotOn(xframe) ;
 
  // Draw all frames on a canvas
  TCanvas* c = new TCanvas("binomialPdf","binomialPdf",400,400) ;
  gPad->SetLeftMargin(0.15) ; xframe->GetYaxis()->SetTitleOffset(1.6) ; xframe->Draw() ;
}

******************************
Output:
[attachment=0]binomialPdf.png[/attachment]

[moneta]

Hi,

You should maybe try to use a binned pdf, such as the RooHistPdf, in this way the integers are taken into accounts as bins

Lorenzo

[tintin2807]

Hi Lorenzo,

Many thanks for your suggestion. It led to the solution. I generated the binomial function using the TH1F constructor, binned it, imported it into RooDataHist() and then passed it on to RooHistPdf() to give me my pdf. It’s a bit fiddly but it does the trick. The key step is to map the function into the TH1F bins.

For those interested see sample code and output below:

#ifndef CINT
#include “RooGlobalFunc.h”
#endif
#include “RooRealVar.h”
#include “RooDataSet.h”
#include “RooDataHist.h”
#include “RooHistPdf.h”
#include “TMath.h”
#include “TCanvas.h”
#include “RooPlot.h”
#include “TAxis.h”
#include “TF1.h”
//
using namespace RooFit ;
//
void Binomial_PdfHistoFit()
{
TH1F *binomialhist = new TH1F(“binomialhist”," TH1F histogram of Binomial function",40,0,41);
Int_t n = 40; // Large N, total no of events
Double_t p = 0.5; // probability
Double_t binomialf[n];
//
for (Int_t i=1; i<n+1; i++)
{
binomialf[i] = TMath::Power(p,i)*TMath::Power((1-p),(n-i))*TMath::Factorial(n)/(TMath::Factorial(i)*TMath::Factorial(n-i));
/////// map entries to bins
binomialhist->SetBinContent(i,binomialf[i]);
}
//// Import histo into RooFit and generate PDF
/////////////////
RooRealVar smalln(“smalln”,“smalln”,0,40) ;
RooDataHist binomialpdfhist(“binomialpdfhist”,“hist”,smalln,Import(binomialhist));
RooHistPdf binomialPdf(“binomialPdf”,“binomialPdf”,smalln,binomialpdfhist) ;
// Create frame and plot
RooPlot xframe3 = smalln.frame(Title(“PDF plot of TH1F generated histogram of binomial function”)) ;
binomialPdf.plotOn(xframe3) ;
TCanvas *c3 = new TCanvas(“c3”,“pdf”,700,350);
c3->Divide(2);
c3->cd(1) ; gPad->SetLeftMargin(0.15) ; binomialhist->GetYaxis()->SetTitleOffset(1.6); binomialhist->Draw() ;
c3->cd(2) ; gPad->SetLeftMargin(0.15) ; xframe3->GetYaxis()->SetTitleOffset(1.6) ; xframe3->Draw();
}