RooChebychev appears to draw the wrong shape

Roofit and RooStats
1

For part of my fit, I use a RooChebychev with one parameter. At the end of the fit, it comes out to be -0.950, which should describe a flat line at -0.950. However, when the fit is drawn it is above 0 and it definitely not constant. What is going on? Code snippets and plot follow.

RooRealVar b0("b0","Background: Chebychev b0",-0.950,-10000,10000); RooArgList bList(b0); RooChebychev bkg("bkg","Background", *mass, bList); [code] RooArgList shapes;
RooArgList yields;
shapes.add(sigL); yields.add(nsigL);
shapes.add(sigS); yields.add(nsigS);
shapes.add(bkgLst); yields.add(nbkgLst);
shapes.add(bkg); yields.add(nbkg);
RooAddPdf totalPdf(“totalPdf”,“totalPdf”,shapes,yields);

RooAbsReal * nll = totalPdf.createNLL(*data,Extended(kTRUE));
RooMinuit m(*nll);
m.setVerbose(kFALSE);
RooFitResult *logfile = m.save(“logfile”,“Log File”);
m.migrad();
m.minos();
m.minos();
[/code]
fit6.pdf (73.5 KB)

2

Hi,

Since one coefficient it is fixed by the normalisation, RooCgebychev with one parameter is a polynomial of degree 1 and not 0, i.e. it is not a constant.

Lorenzo

3

Thank you. Can you tell me how to extract this parameter?

4

Hi,

You pass in the constructor the parameter for RooChebychev (“b0” in your code example). There is no other parameter, since as I said, the offset is fixed by the normalisation.

Lorenzo

5

How can there be no other parameter? It’s still fitting to a polynomial, yes? Which should be described by an equation? I want the equation of the line it’s drawing.

6

You should be able to get the equation from the code implementing it
root.cern.ch/doc/v606/RooChebyc … tml#l00117

Lorenzo

7

Yes, it looks like it does not take a coefficient for the 0th cheby, that is, for 1, but just lets this be determined by the normalization, essentially making the 0th coefficient equal to the normalization factor. 2 questions:

  1. Is there a way for me to call this normalization factor directly?
  2. How does the normalization relate to the other coefficients? Put another way, is the equation norm1+b1T1+b2T2… or is it norm(1+b1T1+b2T2…) where bi are the coefficients and Ti are the polynomials?
8

Hi,

There is no way to get this normalisation factor. It is included in the variable definition of the polynomial, i.e x is redefined to be in the range [-1,1] and the polynomial are defined normalised in that range, as you can see in the code.

Lorenzo

9

I see that the range is mapped onto a different variable from -1 to 1, and I see the note that that is the form in which it is normalized, but you seem to be saying that the redefinition of the variable (that is, mapping it from -1 to 1) is the normalization, which cannot be true, at least not for the way the code redefines the variable. There has to be a normalization factor somewhere.

In any case, is there a place I can request this be changed? This seems needlessly obtuse. It seems better to just be able to add parameters according to the order polynomial you want, as here:

10

bump

11

I suggest you to contact the RooFit author directly for this.