search

SBNSoftware / sbnana

3 stars 14 forks source link

Feature/bckhouse_multisigma_sbn2021c #61

Closed cjbacchus closed 2 years ago

cjbacchus commented 2 years ago

I based this on Jaesung's PR https://github.com/SBNSoftware/sbnana/pull/59 It was helpful for me to go through all the code properly myself and understand what is going on.

I imported the test macro directly, and it is incredibly helpful. Thanks!

For the main weight logic, I removed all the special-casing for the NonRes systematics, including the cuts and extra indices, which simplifies things a lot.

I added direct support to UniverseOracle for looking up a parameter set by name, most probably a multisim one. I think this should allow us to work with files with both genie and flux multisims in them. For looking up individual systematics it still requires that they the parameter set only adjusts a single one. It is possible we will want to expand this one day if we ever start throwing a handful of systematics together and interpolating between the results. This will be necessary to resurrect PredictionLinFit for example (this is an experimental mode of oscillation analysis, not used in the main fit).

Because of the way the weights are stored now, I think it makes sense to only support ISysts for multisigma systematics, and weights for multisim ones.

The output of the test macro looks sensible to me, and I think the same as in the other branch.

It seems like a bad idea for me for the multisim parameter set name to be genie_sbnd_multisim_Genie. I think we should drop the sbnd from that so that we can use the same name for icarus files and simplify oscillation analysis macros.

cjbacchus commented 2 years ago

We will also eventually want this in develop. What is the procedure for that? Does 2021c get merged back there regularly as a matter of course, or...?

miquelnebot commented 2 years ago

We will also eventually want this in develop. What is the procedure for that? Does 2021c get merged back there regularly as a matter of course, or...?

After cutting the production release yes, all get merged to develop as well. If this is "needed" earlier you could duplicate the PR this time for develop.

cjbacchus commented 2 years ago

No, just wanted to make sure I understood the procedure. Thanks!

jedori0228 commented 2 years ago

I updated the GENIE PSet names in https://github.com/SBNSoftware/sbncode/pull/238. Now the knobs are 


Systematic weights in file:
39 parameter sets:
  0: AhtBY_multisigma_Genie, type 3, 6 universes, adjusted parameters:
    AhtBY
  1: BhtBY_multisigma_Genie, type 3, 6 universes, adjusted parameters:
    BhtBY
...  38: multisim_Genie, type 0, 10 universes, adjusted parameters:
    AhtBY
    BhtBY
    CV1uBY
    CV2uBY
...```