I am building my signal model with HistFactory. I have a signal sample and two background samples.
The two background samples have yields constrained by external measurements which I implemented as OverallSys.
I first use the model to fit a toy data sample and then use it to get an asymptotic limit with RooStats. So far everything works as expected.
Issues arise when I want to set a systematic on the efficiency in the signal yield where the efficiency is one of the normalisation factors for this sample. I naively tried to add the systematic as overall systematic, as for the background. The fit to the toy looks good, but when I feed the model to Asymptotic limit calculator I get a nonsense result.
Is this the right way to implement systematics on the signal sample?