EvalBackend for ProfileLikelihoodCalculator

Hi experts,

Is there a way to set the EvalBackend (in the way it is done for createNLL) for the ProfileLikelihoodCalculator?

For context, I am trying to use a ProfileLikelihoodCalculator with a morphed PDF, and when doing so get a lot of RooAddPdf::fixCoefNormalization(nset) warnings. I noticed in the morphing tutorial there is a note in the TODO that this is caused by the new CPU backend, and a workaround is to use the legacy with the NLL. I also don’t want to just ignore the warning - when I do a fit with NLL (with the legacy backend setting) I get a sensible result, but running with the ProfileLikelihoodCalculator I get something that doesn’t converge properly.

Thanks in advance.

Hi @zurowski,

thank you for your question. I’ll add @jonas for further discussion.

Cheers,

Marta