Transcript Zprime_150411

Data driven estimate of electron fake background
with fake rate measured from jet data
Xin Wu
University of Geneva
1
Method and applcation
•
“Fakeable” (“denominator”) object: a jet, or better a jet reconstructed
as an electron object, passed some loose identification cut
•
Fake rate: probability of a fakeable object to pass the nominal electron
ID cuts (eg. isEM Medium + BL cut)
•
Estimate fake rate from jet samples
– No electron trigger bias -> flexibility in defining fakeable selection
– Real electrons need to be removed or subtracted (eg. Z, W, …)
– Statistics is often a problem because of prescales
fake rate =
# of fakeable objects passed nominal ID cut - # of real electron
# of fakeable objects - # of real electron
• often as function of pt and eta
•
Application for Z’ analysis : electron+fake background (eg. W+jets)
– Highest pt electron passed nominal cut pairs with all fakeable
objects in the event, the fake rate of the fakeable object is the
event weight of the pair. The sample of all the pairs is the
electron+fake background
– More elaborated paring schemes possible
2
Fakeable selection
•
Minimize bias (differences) between fakeable objects in the jet sample
and in the analysis sample but retain enough statistics
– In particular jet trigger threshold bias
•
Basic selection A (Author electron with nominal track cut)
– same grl, vtx, eta, OTX
– Track cut in Medium (TRACKINGNOBLAYER)
– B Layer cut
•
Further selections
– R: Hadronic leakage (HADLEAKETA_ELECTRON)
– L: Hadronic leakage + Reta (isEM Loose)
– (can consider add loose isolation, to R, if isolation used in the
nominal selection)
•
Validation
– Invariant mass low mass region
3
Data sample
•
JetTauEtmiss AOD from run 178044 and 178109, ~13 pb-1
– data11_7TeV.00178044.physics_JetTauEtmiss.merge.AOD.f354_m765
– data11_7TeV.00178109.physics_JetTauEtmiss.merge.AOD.f354_m765
– Egamma GRL
– Z removal: reject events containing a pair of the fakeable objects
of base selection with invariant mass between 70 and 110 GeV, if
one of the object also passed isEM loose cut (0.4% event rejected)
events
•
Selection A, Et > 25GeV:
•
Selection R (A + HadLeakage):
•
Selection L (A + Loose) :
4,035 4,055
•
Medium selection:
1,410
•
Jet trigger considered : EF_jXX_a4_EFFS
– XX: 20,
L1
365,474
objects
579,894
19,008 19,274
1,419
30,
40
55
75
100
135
RD0 J10
J15
J30
J50
J75
J75
4
Z veto
5
Mee with two Loose+BL
•
Answer to the question raised last week
– no statistics of loose-loose pairs from the jet sample
6
Fakeable object Et distribution, selection A
7
Fakeable object Et distribution, selection R
•
Would be nice to has x2 less prescale on EF_j100
8
Fakeable object Et distribution, selection L
•
Would help to have lower prescales on EF_j100 and EF_j75
9
Et distribution, nominal Medium selection
•
Real electrons from W not yet removed/subtracted
10
Fake rate, selection A
•
Clear bias from trigger threshold
•
W not yet removed subtracted
11
Fake rate, selection R
•
Trigger bias and W need to be removed
12
Fake rate, selection L
•
Limited statistics
•
Trigger bias and W need to be removed
13
Et of fakeable objects in Zprime sample
•
14
Conclusion
•
Preliminary studies of extracting electron fake rate from jet data
•
3 sets of fakeable selections
– Basic selection (A): author electron with nominal track cut
– Selection R: A + Hadronic leakage
– Selection L: A + isEM Loose
• Selection L has limited statistics for fake rate calculation
•
Z removal with invariant mass cut, W removal to be done
– W can also subtracted using MC, but luminosity with prescales of
many jet samples would be complicated
•
Important trigger bias need to be removed
– Will study removing fakeable objects matched to the trigger jet
•
Could profit from lower prescale for jet samples, in particular EF_j100
and EF_j75
15