P.d.f normalization integral is zero or negative

Roofit and RooStats
1

Hi,
I’m trying to plot a profile likelihood interval using:
ProfileLikelihoodCalculator pl(*data,mc);
pl.SetConfidenceLevel(0.95);
LikelihoodInterval interval = pl.GetInterval();
LikelihoodIntervalPlot plot(interval);
plot.SetNPoints(50);
plot.Draw("");

When I do, I get a series of errors like this:
[#0] WARNING:Minization – RooMinimizerFcn: Minimized function has error status.
Returning maximum FCN so far (26831.6) to force MIGRAD to back out of this region. Error log follows
Parameter values: f=0.418067, mGau=0.665142, mLan=178.942, mu=0.3775, wG=1.53094, wL=34.3664
RooGaussian::SignalG[ x=Mjj mean=mGau sigma=wG ]
p.d.f normalization integral is zero or negative @ x=Mjj=44.3092, mean=mGau=0.665142 +/- 0.945268, sigma=wG=1.53094 +/- 1.15983
And so on…

Which I don’t understand. The PDF (a gaussian plus two landaus) is well behaved and the fit to data looks good. I guess I’m missing something.

Any suggestions would be welcome.

2

From the log I see you are having a Gaussian with x=44 and mu=0.66 and sigma=1.53. Are you sure the range of the Gaussian is not off the peak ?
This could explain the problem

Lorenzo

3

Thanks for the quick response. The Gaussian is set up with a mean of 38.32:
RooRealVar mGau(“mGau”,“Mean of Gaussian”,38.32,0,100) ;

And the signal+background fit looks just fine (should by definition since the toy MC data is generated with the same model)

I think it’s LikelihoodIntervalPlot that’s scanning over different values of mGau etc.

4

Yes it is correct that the LikelihoodINtervalPlot is scanning several parameter values. Which one is your parameter of interest (POI) ?
You can also set the POI range using LikelihoodIntervalPlot::SetRange(x1,x2)

Lorenzo