Constrain option in Roofit


I’ve been trying to understand the point/behaviour of the Constrain option in Roofit… Specifically, if I have a simple model and I add to it a Gaussian constraint term…lets say, constraining the background estimate in a simple counting experiment…

p(n;s+b) = Pois(n|s+b)Gaus(b|b0,delta_b)

then if I fit my model to by data (the n_observed), it doesn’t seem to matter whether I do “Constrain(b)” or not in my fitTo call. If I leave it out, I see from RootFit output messages that the b var gets constrained anyway.

In the case of multiple gaussian constraint terms, if I use “Constrain(…)” option, then it seems (from output messages) to only apply constraints to the terms I specify and I get different fit results… but does this mean that the gaussian constraint terms I added to my model are not included? Is that the only point of the Constrain option? To remove gaussian constraints from it?