Comments (6)
Regarding b-tagging SF, the conclusions from today's meeting is the following:
- it's fine to use jets with nominal JER corrections in the evaluation of the b-tagging SF;
- the BTV POG supports the usage of b-tagging SF computed for V1 of reduced set of JES uncertainties that were circulated about half a year ago. There will be no b-tagging SF computed for V2 of reduced set of JES uncertainties (at least for the Moriond time scale) because the effect on b-tagging SF is expected to be small (relative to V1);
- since the b-tagging SF are found to respond differently in case of tune CUETP8M1 compared to CP5 tune, we're expected to have dedicated measurement of the b-tagging SF specifically for the CUETP8M1 tune in 2 weeks.
What this basically means to us is that we can post-process 2017 and 2018 MC samples to precompute b-tagging SF for the reduced JES sources as soon as possible, but we would have to postpone the post-processing of 2016 MC samples until the b-tagging SF for the CUETP8M1 tune are made available (because 2017 and 2018 samples are produced with the CP5 tune).
The post-production itself needs to be rearranged such that the b-tagging SF module runs after the JetMET module that prepare the energy-shifted jet collections. The b-tagging SF module has to be run separately from other modules because it cannot access the jet energy collections created in the JetMET module if both modules are run simultaneously. This will certainly increase the runtime per event that is spent on post-production.
The third aspect of this whole ordeal is that we don't have infinite disk space at our disposal: once we have new post-processed Ntuples available, the Ntuples we're currently using have to be deleted. If we could drop the reduced set of AK8 systematics, we could reduce the time spent on post-production, as well as the disk space required to store the Ntuples.
from hh-bbww.
I tried to understand the origin of 10% uncertainty associated with the b-jet energy regression and came across HIG-18-027 and its complementary AN2018/092v3. It's not entirely clear to me if we're applying the b-jet energy corrections as documented in the AN. For instance, lines 182-184 and 296-297 imply that before applying the b-jet energy corrections, the nominal JER corrections should be undone, and a SF 1.1 +/- 0.1 applied. I suspect that the +/- 0.1 part refers to the 10% uncertainty mentioned in the HH->4b AN. It may also be that the JetMET POG didn't provide JER SF at the time of publishing the paper, and the whole discussion about JER SF has become outdated.
The second item is that the b-jet energy regression corrections are currently applied only to pT of the jet:
Particle::LorentzVector
RecoJet::p4_bRegCorr() const
{
return math::PtEtaPhiMLorentzVector(this->pt()*bRegCorr_, this->eta(), this->phi(), this->mass());
}
and not to both pT and mass of the jet:
Particle::LorentzVector
RecoJet::p4_bRegCorr() const
{
return math::PtEtaPhiMLorentzVector(this->pt()*bRegCorr_, this->eta(), this->phi(), this->mass()*bRegCorr_);
}
Also, I just learned that we're using an older b-jet energy regression training. The training was updated with NanoAODv7. It also looks like the official NanoAOD is misconfigured to use 2018 training for all eras, even though there are separate trainings for each era. I'll probably shoot an email to XPOG at some point.
from hh-bbww.
Since we still don't have b-tagging SF for the older 2016 tune, I'm temporarily assuming that all 2016 samples are produced with the CP5 tune (link to code):
'isTuneCP5' : (self.era == "2016"),# and 'TuneCP5' in sample_name), #TODO change it back
from hh-bbww.
The event yields are compatible with what we had before in 2016 ggF NLO HH->bbWW DL sample. I'll also update the b-tagging SF that we need for the old 2016 tune.
from hh-bbww.
WW renromalization/factorization scale uncertainties are missing -> will add them shortly.
from hh-bbww.
Looks like the verdict is to not apply the 10% uncertainty that is associated with b-jet energy resolution, and to stick with the JER uncertainties provided by the JetMET POG.
from hh-bbww.
Related Issues (20)
- Add ttH, H->bb + enable a few more samples
- Skim samples for bb1l analysis HOT 2
- Make LBN work with CMSSW HOT 3
- Prototype to 4-jet assignment HOT 1
- New signal samples HOT 3
- Relaxed loose lepton definition in DL analysis HOT 2
- Restrict the list of syst unc in SL and DL HOT 2
- Update DL sample cross sections HOT 1
- Reskim samples with the relaxed lepton definition HOT 2
- Also use VBF jets to compute the PU jet ID sF HOT 6
- Status flag of LHE particles missing HOT 3
- Optimize the hadd step
- Updates regarding samples
- b-jets missing PU jet ID cuts HOT 1
- Script for extracting relative MC closure uncertainties HOT 1
- Encode LBN category name in addSystFakeRate cfgs HOT 1
- Revisit reweighting in non-resonant analysis HOT 1
- Missing systematics HOT 10
- Reweigh down 2016 resonant HH events that have W->tau nu HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from hh-bbww.