Statistical error on Number of signal events

Roofit and RooStats
1

Hi,

I realised that in some cases after doing a likelihood fit and extracting the number of signal events, I find the error to be way bigger than the actual number of signal events. (As you can see in the photo attached)

test
test.png1198×947 18.4 KB

I’m doing a Gaussian fit and I let the parameters float. For a J/psi fit :

  RooRealVar mean1("#mu","mean of gaussians",3.097,3.096,3.098);
  RooRealVar sigma1("#sigma","width of gaussians",0.0048,0.003,0.005);
  RooGaussian sig1("sig1", "sig1", x, mean1, sigma1);
  RooRealVar a0("a0","a0",0.0005,-2.,1.) ;

Could you please let me know if I’m doing something wrong that contributes to this big error on the number of signal events ? Can I avoid it?

2

In the figure it looks like the number of events is divided by a very small number. What is in your y-axis? Isn’t this what’s giving you the large errors? How does Nsig relate to that small number?