AddOverallSys and lognormal errors

[Bonn]

Hi there,
Im building a statistical model using HistFactory and need to incorporate a normaltisation factor whith its error. The factor itself is implemented using Sample.AddNormFactor.
As far as I can tell the error on the normalisation factor is supposed to be modeled by a lognormal since it’s strictly positive. How do i do that with HistFactory?
Sample.AddOverallSys only introduces a gaussian constraint, right?

Cheers

[eguiraud]

Hi @Bonn ,
sorry for the high latency, let’s ping our RooFit experts @moneta @jonas .

Cheers,
Enrico

[moneta]

Hi,

Sorry for the late reply. In the HistFactory (as described in the Users guide), the Normalisation Factor will not be un unconstrained parameter, so not associated with any constraint and
the OverallSys with a Gaussian constraint.
However, by using an exponential interpolation for the OverallSys parameter, one can have the equivalent of a log-normal constraint.

Cheers

Lorenzo