search

GENIE-MC / Reweight

The GENIE Reweight product includes a collection of tools for propagating model uncertainties.
3 stars 19 forks source link

Fixing GReWeightResonanceDecay Double Application #34

Closed leehagaman closed 1 month ago

leehagaman commented 1 month ago

@nitish-nayak and I found an issue that causes the GReWeightResonanceDecay calculator to be applied twice, leading to squared weights. res_dk and hadro_res_decay both apply the same GReWeightResonanceDecay calculator, and here I delete the res_dk line. This was first noticed after seeing larger than expected uncertainty for NC Delta radiative decays in MicroBooNE. The three knobs affected are BR1gamma, BR1eta, and Theta_Delta2Npi.

It seems like this code was copied, so the issue also exists in codebases for MicroBooNE, SBN, ANNIE, and DUNE.

sjgardiner commented 1 month ago

Great catch @leehagaman and @nitish-nayak. Many thanks for making this fix!

tomjunk commented 2 weeks ago

The DUNE link above points on an old, unmaintained DOxygen instance which also ran on LArSoft. So the GenieWeightCalc pointed to by the link that says DUNE above is in fact the one in larsim. This is addressed in https://github.com/LArSoft/larsim/pull/142

Thanks for spotting this! Steven should know more about the knock-on effects of changing things for people who are currently running the buggy version and may get a mixture of buggy and non-buggy weights. Though they should use fixe releases if they need bug-for-bug compatibility, so I am almost always in favor of just fixing bugs and moving forwards.