roofit resolution/convolution

From: Anne Keune <anne.keune_at_cern.ch>
Date: Wed, 26 Jan 2011 15:33:04 +0100

Dear all,

I have a question concerning the convolution with a bifurcated Gaussian resolution function. I am studying a lifetime distribution, which is by definition positive. Therefore, close to 0, the spread caused by the resolution is only towards the right, i.e. towards the higher lifetime values, because it is impossible to go to negative lifetime values. As a resolution function, I therefore thought it would be a good idea to use a bifurcated Gaussian, with its right sigma a constant, but its left sigma a function.

  ///RESOLUTION MODEL
  RooRealVar life("life","life",0,0.002); //my variable                                                                                                                    

  RooRealVar mean("mean","mean gaussian",0);   RooRealVar sigma1("sigma1","sigma1 gaussian",0.0001);

  RooRealVar a3("a3","a3",0.5);
  RooRealVar c3("c3","c3",0.0001);
  RooRealVar d3("d3","d3",0.0002);

  RooFormulaVar sigma2("sigma2","(life<d3)*(life-d3)*a3+c3",RooArgList(a3,c3,d3,life)); //doesn't work   RooBifurGauss gauss_model("gauss_model","gauss_model",life,mean,sigma2,sigma1);

I would like to use this is a RooFFTConvPdf, but get the error message: [#0] ERROR:Eval -- RooFormula::eval(sigma2_fft): Formula doesn't compile: (life<d3)*(life-d3)*a3+c3

If I use the bifurcated Gaussian without the RooFFTConvPdf, then it does compile. Am I not allowed to use a RooBifurGauss in a RooFFTConvPdf? Is there another way of doing this?

Cheers,

Anne Received on Wed Jan 26 2011 - 15:33:10 CET