Hi @David_Vannerom,
thanks for asking your question in the ROOT forum!
My question is: how do I get the upper limit (and corresponding +/- 1 and 2 sigma bands) as a function of the HNL mass?
Looks like you made your plot by adapting the StandardHypoTestInvDemo. You see your expected and observed upper limits are where the p-value is 0.05, as indicated by the red line. I assume your Brazilian plot was made with the mass fixed to some specific value, right? To get the limit as a function of the mass, you have to run the StandardHypoTestInvDemo multiple times with different values for the mass, as also explained in this forum post.
I don’t understand what this plot tells me, why do I only get the p-value as a function of a single scalar (nsig) and not a binned histogram of the HNL mass?
I can’t see your code, but most likely you have done the plot fixing the HNL mass to some specific value. The plot shows the p-value for your test statistic (should be profile Likelihood one sided) computed with the asymptotic formulae as explained in this reference (page 36 in the pdf, the p-value is alpha in the formula). In particular, the plot tells you that your median expected upper limit is roughly 8 times the standard model prediction (ready by eye from the plot, the actual numeric values can be found in the standard output text), and the distribution of the upper limit is given by the uncertainty band. As explained above, to get limits as a function of the HNL mass you have to run this procedure multiple times for different values of the mass.
By the way, you’re saying “these are my signal and BG models”, but I only see a single histogram. What’s that? Signal, background, or both?
I hope my answer helps you, if you still have further questions or if something is unclear, don’t hesitate to ask!
Cheers,
Jonas