Simultaneous fit to a binned dataset in RooFit

Roofit and RooStats
1

Dear All,

I am trying to fit a simultaneous PDF to a binned dataset : there is one fit variable (the mass) and a bunch of categories reflecting different datataking periods, magnet polarities, decay modes, etc. I first make an unbinned dataset containing all the events and then bin it

RooDataHist binneddata("binneddata","binneddata",RooArgSet(qval,rswscat,compptcat,btaucat,channelcat,magpolcat,yearcat,bflavcat),data);

“qval” is the mass variable, the rest are categories. The eventual PDF contains various functions of the mass variable, whose parameters I want to variously split and/or share depending on the values of the various categories. All events have the qval and the various categories set when I create the dataset.

Then I try to make a simultaneous PDF (note the following is python code)

SimPdfTool = RooSimWSTool(fitworkspace)
SimWSPDF   = SimPdfTool.build("SimWSPDF","WSPDF",RooFit.SplitParam(varstosplit,"compptcat,yearcat,channelcat"))

I don’t split in the simultaneous fit on all the categories. This gives the following type of error once I try to fit

[#0] ERROR:InputArguments -- RooCategory::isStateInRange(rswscat) ERROR: must specificy valid range name and state name [#0] ERROR:InputArguments -- RooCategory::isStateInRange(btaucat) ERROR: must specificy valid range name and state name [#0] ERROR:InputArguments -- RooCategory::isStateInRange(magpolcat) ERROR: must specificy valid range name and state name [#0] ERROR:InputArguments -- RooCategory::isStateInRange(bflavcat) ERROR: must specificy valid range name and state name

I have tried modifying the tutorial related to the RooSimWSTool, and I can reproduce the problem in a toy case : if I bin the dataset the problem is there, and if I do the same thing with the unbinned dataset then the problem is not there.

I would be grateful if someone could explain where I am going wrong. I am using ROOT v5.34/04. Doing an unbinned fit is not an option because of the size of the dataset.

Many thanks,

Vava

2

Hello,

in case it helps, here is the modified tutorial script which reproduces the problem. If you replace “binneddata” by “combData” in the “fitTo” line, the fit works fine.

Thanks,

Vava

/////////////////////////////////////////////////////////////////////////
//
// 'ORGANIZATION AND SIMULTANEOUS FITS' RooFit tutorial macro #504
// 
// Using RooSimWSTool to construct a simultaneous p.d.f that is built
// of variations of an input p.d.f
//
//
// 07/2008 - Wouter Verkerke 
//
/////////////////////////////////////////////////////////////////////////

#ifndef __CINT__
#include "RooGlobalFunc.h"
#endif
#include "RooRealVar.h"
#include "RooCategory.h"
#include "RooDataSet.h"
#include "RooGaussian.h"
#include "RooConstVar.h"
#include "RooPolynomial.h"
#include "RooSimultaneous.h"
#include "RooAddPdf.h"
#include "RooWorkspace.h"
#include "RooSimWSTool.h"
#include "RooPlot.h"
#include "TCanvas.h"
#include "TAxis.h"
#include "TFile.h"
#include "TH1.h"
using namespace RooFit ;


void rf504_simwstool()
{
  // C r e a t e   m a s t e r   p d f 
  // ---------------------------------
  
  RooRealVar x("x","x",-8,8) ;

  // Construct signal pdf
  RooRealVar mean("mean","mean",0,-8,8) ;
  RooRealVar sigma("sigma","sigma",0.3,0.1,10) ;
  RooGaussian gx("gx","gx",x,mean,sigma) ;

  // Construct background pdf
  RooRealVar a0("a0","a0",-0.1,-1,1) ;
  RooRealVar a1("a1","a1",0.004,-1,1) ;
  RooChebychev px("px","px",x,RooArgSet(a0,a1)) ;

  // Construct composite pdf
  RooRealVar f("f","f",0.2,0.,1.) ;
  RooAddPdf model("model","model",RooArgList(gx,px),f) ;

  // C r e a t e   m o d e l   f o r   c o n t r o l   s a m p l e
  // --------------------------------------------------------------

  // Construct signal pdf. 
  // NOTE that sigma is shared with the signal sample model
  RooRealVar mean_ctl("mean_ctl","mean_ctl",-3,-8,8) ;
  RooGaussian gx_ctl("gx_ctl","gx_ctl",x,mean_ctl,sigma) ;

  // Construct the background pdf
  RooRealVar a0_ctl("a0_ctl","a0_ctl",-0.1,-1,1) ;
  RooRealVar a1_ctl("a1_ctl","a1_ctl",0.5,-0.1,1) ;
  RooChebychev px_ctl("px_ctl","px_ctl",x,RooArgSet(a0_ctl,a1_ctl)) ;

  // Construct the composite model
  RooRealVar f_ctl("f_ctl","f_ctl",0.5,0.,1.) ;
  RooAddPdf model_ctl("model_ctl","model_ctl",RooArgList(gx_ctl,px_ctl),f_ctl) ;

  // G e n e r a t e   e v e n t s   f o r   b o t h   s a m p l e s 
  // ---------------------------------------------------------------

  // C r e a t e   c a t e g o r y   o b s e r v a b l e s   f o r   s p l i t t i n g
  // ----------------------------------------------------------------------------------

  // Define two categories that can be used for splitting
  RooCategory c("c","c") ;
  c.defineType("run1",1) ;
  c.defineType("run2",2) ;

  RooCategory d("d","d") ;
  d.defineType("foo",1) ;
  d.defineType("bar",2) ;


  // Generate 1000 events in x and y from model
  RooDataSet *data = model.generate(RooArgSet(x),100) ;
  RooDataSet *data_ctl = model_ctl.generate(RooArgSet(x),2000) ;

  // S e t u p   S i m W S T o o l 
  // -----------------------------

  // Import ingredients in a workspace
  RooWorkspace w("w","w") ;
  w.import(RooArgSet(model,c,d)) ;

  // Make Sim builder tool
  RooSimWSTool sct(w) ;

  // B u i l d   a   s i m u l t a n e o u s   m o d e l   w i t h   p r o d u c t   s p l i t
  // -----------------------------------------------------------------------------------------

  // Build another simultaneous p.d.f using a composite split in states c X d
  RooSimultaneous* model_sim2 = sct.build("model_sim2","model",SplitParam("mean,a1,a0,f","d")) ;

  // Print tree structure of this model
  model_sim2->Print("t") ;

  RooDataSet combData("combData","combined data",x,Index(d),Import("foo",*data),Import("bar",*data_ctl)) ;

  RooDataHist binneddata("binneddata","binneddata",RooArgSet(x,c,d),combData);  

  model_sim2->fitTo(binneddata);

}
3

Dear all,

I have the same problem. The fit works and I do not get these “RooCategory::isStateInRange(tag) ERROR” messages when I switch to an unbinned fit. However, also for me this is not a real option given the size of my data set. Any help is appreciated.

Thanks
Jeroen

4

Dear all,

The bug is fixed by Wouter on the 22th of March. When I get the latest patch (svn co root.cern.ch/svn/root/branches/v5-34-00-patches) and compile it the problem is fixed and I can happily fit to binned data sets.

Cheers
Jeroen

5

Since I am also hit by this bug (while trying to reduce a dataset with roodatahists) the actual patch is at root.cern.ch/viewvc/branches/v5- … 2&r2=48913