I am trying to use histfactory to build the systematic uncertainties in my analysis.
I am implementing the .xml example that I found by typing prepareHistFactory.
In the background sample I used:
to build the systematics related to the +/- 1 sigma change in the shape.
I use the model built by HistFactory to calculate the upper limit on the signal cross section. It seems to work.
However, I would like to see how the results on the upper limit calculation change for different values of the interpolation code. I read that the default one is 0 (Piecewise Linear).
How can I do?
Dear all,
I have another question for the experts, regarding almost the same topic, so I will write it here.
I would like to put together the shape and normalization factor systematic in histfactory.
I have to calculate the impact of a systematic uncertainty that affects both the shape and the normalization of my Signal. I implement it in HistFactory as follows:
Where I give the Signal, that is the nominal histogram, normalized to the nominal number of events.
Then, I find +29% of events for a +1 sigma variation of the systematic, and the signal shape is called “Signal_p1s”. I gave this histogram normalised to 1.29 times the nominal normalization of the Signal (and here is my doubt).
In the same way I calculated also Signal_m1s, that is normalised to 0.74 times the nominal normalization of the Signal.
Then I used a single nuisance parameter for the two, since I expect that, for example, when the nuisance parameter syst1 is +(-) 1, I will have that the signal has Signal_p1s (Signal_m1s) shape, with an integral that is 1.29 (0.74) times the nominal one.
I would like to understand if I am doing it correctly, in particular for what concerns the histograms normalization. Have I to give them normalized to the same area or is it correct as I am doing? Am I in this way double counting the systematic?