RooFit, formulas in RooHistPdf

ROOT
1

Hi,

I want to create a RooHistPdf from an observable in a dataset. The twist is that I don’t want to use the observable as it is, but apply a formula to it, but I can’t seem to do this since a RooFormulaVar isn’t of “fundamental” type. This code snippet reproduces my problem:

{
using namespace RooFitShortHand;
RooRealVar sigma(“sigma”,“sigma”,0.0,0.2);
RooLandau sigma_dist(“sigma_dist”,“sigma_dist”,sigma, C(0.05),C(0.01));
RooDataSet *datasigma = sigma_dist.generate(RooArgSet(sigma),5000);
RooFormulaVar d1(“d1”,“sqrt(@0)”,RooArgSet(sigma));
RooDataHist h1(“h1”,“h1”,d1,*datasigma);
RooHistPdf (“h1pdf”,“h1pdf”,d1,h1);
}

which gives me the error
RooAbsSet::initialize(h1): Data set cannot contain non-fundamental types, ignoring d1

Does anyone know what I am doing wrong here? Is there another way to do this?

cheers
Aras

1 Like
2

Hi Aras,

Sorry for the late reply. I was on vacation.

Datasets can only contain fundamentals (i.e variables), not observables.

If you want to make a plot of a distribution of a derived observable,
you can do that easily, but it requires an intermediate step: you add a column
to the dataset with the calculated values of your transformation function on
the data points already contained in the dataset. Here is a modified version
of your example:

using namespace RooFitShortHand;
RooRealVar sigma(“sigma”,“sigma”,0.0,0.2);
RooLandau sigma_dist(“sigma_dist”,“sigma_dist”,sigma, C(0.05),C(0.01));
RooDataSet *datasigma = sigma_dist.generate(RooArgSet(sigma),5000);
RooFormulaVar d1_func(“d1”,“sqrt(@0)”,RooArgSet(sigma));

RooRealVar* d1 = (RooRealVar*) datasigma->addColumn(d1_func) ; // <–

RooDataHist h1(“h1”,“h1”,*d1,*datasigma);
RooHistPdf (“h1pdf”,“h1pdf”,*d1,h1);

The magic line is the one marked with the arrow, which adds a column to your
dataset with the output of the transformation that you give it.

I’ll answer your other posting on conditional p.d.f.s tomorrow.

Wouter

3

Thanks, this seems to work!

cheer
Aras