aboletti commented 2 years ago

This issue will be used to track the configuration of the reweighting (like input variables and hyperparameters) of the latest setup, but also of some intermediate version that are used to test the fit. The version naming can be revisited, but I propose:

v0 for the configuration tested before the pre-approval
vX for any configuration that is used to produce efficiencies and test the fit (while we can use subversion vX.Y if we want to distinguish different tests for which we check only the training result and distributions)
vX-"name" in case there are different variations of the configurations released at the same time (like v1-daug-pT and v1-noDaug-pT, otherwise we can just give a letter v1-a and v1-b, and we report here the meaning)

V1 configuration (10 March)

include bug fix to the BDT background sample
B pT and eta used as input, no daughter pT used
replace all significances with value and error input variables
only 2018
efficiency and MC mass fit results in: /eos/user/a/aboletti/BdToKstarMuMu/eff-KDE-theta-v4-XGBv1/

sarafiorendi commented 2 years ago

ntuples with the bug fix for the background sample are being copied here /eos/cms/store/user/fiorendi/p5prime/2018/skims/newphi/fixBkg

aboletti commented 2 years ago

Thanks! @XuelongQin can you produce the weights based on these 2018 ntuples and the input variables described above?

XuelongQin commented 2 years ago

Okay, but I am a little confused about the second configuration. I guess it means that I should generate two versions, one version with only bPt, the other with bPt, trkPt and muon Pt?

aboletti commented 2 years ago

Yes, for sure we need the version with bPt, but no trk or mu pT. If you have time, it could be interesting to have also a version with the trk and mu pT as well, for a quick comparison, but it's not critical

XuelongQin commented 2 years ago

For the V1 configuration: You can find the "odd" parity ntuple with data/MC weights (the branch named "MCw") in /eos/user/x/xuqin/workdir/B0KstMuMu/reweight/Tree/final/gitv1/Foreff/ And you can find the plots of comparison of before/after reweighting in /eos/user/x/xuqin/workdir/B0KstMuMu/reweight/Tree/final/gitv1/Plots/

aboletti commented 2 years ago

Thanks, @XuelongQin. I'm sending the efficiency production. Can you push the exact code used to produce them and create a git label marking it as "v1", so that we have a reference for the future?

aboletti commented 2 years ago

@XuelongQin in addition to the previous request, can you also produce the MC ntuple with weighted events in the mass range 4.9-5.7 GeV? They should not be used in the training (so the v1 training is still OK), but they (and their weights) are needed to fit the WT mass shape.

sarafiorendi commented 2 years ago

@XuelongQin can you also add to your last set of slides the eta distributions of the final state particles? (I'm just asking because I want to compare to my new data/MC plots). then we can maybe include here a link to the full set of plots so that we find them quickly. thanks!

XuelongQin commented 2 years ago

Sorry for the delay, I didn't check this issue these days, I will do these things tommorrow.

XuelongQin commented 2 years ago

@sarafiorendi You can find the plots in the link I gave above: /eos/user/x/xuqin/workdir/B0KstMuMu/reweight/Tree/final/gitv1/Plots/ I included the eta distributions of all final particles there.

XuelongQin commented 2 years ago

@aboletti You can find the ntuple named "recoMCDataset_rw_b*_2018_p1_49_57.root" in the mass range 4.9-5.7 in the link: /eos/user/x/xuqin/workdir/B0KstMuMu/reweight/Tree/final/gitv1/Foreff/ And you can check the code of version 1.

CMSKStarMuMu / DataMCreweighting

Reweighting configuration #1

V1 configuration (10 March)