I’m trying to make a fit with HistFactory, where the POIs of our measurement have normalization and shape variations simultaneously.
What we would like to use in the fit are unconstrained Nuisance Parameter which behaves like HistoSys, i.e. nuisance parameters with shape and normalization variations, but for which no penalty is paid in the likelihood function when they are varied.
Apparently this is not possible, so we implemented them as HistoSys and use large variations as 1 sigma template. The 1 sigma template variation is >> of what we expect to fit in order to minimize the penalty in the likelihood.
Is it possible to use a really unconstrained HistoSys instead?