Transcript No. of Btagged jets - Indico

Data-driven Method for W+jets Estimation
• Introduction
• Matrix Construction
• Estimation
• Conclusion
Xiaohu Sun, Shandong Univ.
Julien Donini, LPSC
1
Goal
Introduction
Matrix Construction
Estimation
Conclusion
• To find out how many
W+jets events passed
the single top T channel
event selection
Distribution after event selection (Btag)
– Estimate the total No. of Btagged jets in W+jets events
– Convert the No. of jets to No. of events
• Tag rate matrix method
– Used in CDF ttbar x-section measurements (ex: PRL 96,
202002 (2006), G. Cortiana et al.)
2
Tag Rate Matrix Method
Introduction
Matrix Construction
Estimation
Conclusion
Tag Rate Matrix:
– use W+jets events only
– find some promising variables sensitive to b-tagging efficiency
– parametrization of per-jet tagging probability from 3-jets
events
– apply it to 2-jets events (like extrapolation)
=
X
Btag efficiency over jet Pt
of event W+jets
2jet events
2jet events
3jet events
Jet distribution over jet Pt
of event W+jets
Estimated jet No. of W+jets
3
Introduction
Matrix Construction
Estimation
Conclusion
Data
• D3PD (release 14)
– https://twiki.cern.ch/twiki/bin/view/AtlasProtected/ListOfSingleTo
pD3PD#I_1_Ntuples_without_overlap_remo
Data are normalized to 200 pb-1
4
Calculate the efficiency

N
jet
tagged
jet
total
N
Introduction
Matrix Construction
Estimation
Conclusion
Tagged jets Efficiency
(All numbers are jet numbers)
Tagged means –ln(JetProbWeight)>4.26
Simple example
• We calculate the b-tagging jets eff(W+3j) for 3-jet events, and apply it to
estimate the total number of b-tagged jet(W+2j) for 2-jets events:
– eff(W+3j)=1.57%+-0.06%
– N (2-jet) tagged,estimate= Ntotal(2-jet)*eff(W+3jets)= 2002+-76
• But this doesn’t work !
• the observed b-tagging jet eff(W+2j): eff(W+2j)=1.47%+-0.03%
• And real number of tagged jets in 2-jet events is:
– N (2-jet) tagged,observe= 1876+-28
5
So, Tag Rate Matrix Method
Introduction
Matrix Construction
Estimation
Conclusion
We are trying to find out variables which are sensitive to
the efficiency described as below
 (x) 
N
jet
tagged
jet
total
N
( x)
( x)
b-tagging Efficiency
(All numbers are jet numbers)
Tagged means –ln(JetProbWeight)>4.26
X is a vector of candidate variables
x  ( PT jet , jet , cos  jet , ET , cos  jet , Lep , ET cos  jet , ET , PTLep cos  jet , Lep )
6
Introduction
Matrix Construction
Estimation
Conclusion
Event selection
• Events are required to pass the usual event preselection (see ATL-COM-PHYS-2009-572)
• Additional cuts (upper limitation) are applied when the
matrix is constructed.
• So we lose some jets.
Cuts
Jet lose on W+2j
Jet lose on W+3j
0.78%+-0.02%
1.88%+-0.06%
|  jet | 2.50
| ET cos  jet , ET | 250.0GeV
13.67%+-0.10%
12.84%+-0.16%
0.16% +- 0.01%
0.36% +- 0.03%
| PTLep cos  jet , Lep | 250.0GeV
0.08% +- 0.01%
0.16% +- 0.02%
PT jet  250.0GeV
NO jetProb weight info in the eta range beyond 2.5
7
Promising variables
 W  2 j   W 3 j
 W 2 j
8
9
Promising variables
10
Promising variables
11
Promising variables
Estimated No. of btagged jets
Introduction
Matrix Construction
Estimation
Conclusion
Estimated No. of btagged jet for W+2j events
based on 2-dimensional eff matrix from W+3j events
Variables
Estimated No. of btagged jet
{PT jet , jet }
1894 +- 44
{PTLep  cos  ( j , Lep)，
cos  ( j , ET )}
1986 +- 46
{PT jet , cos  ( j , ET )}
1906 +- 45
{PT jet , PTLep  cos  ( j , Lep)}
1903 +- 44
{ jet , cos  ( j , ET )}
1954 +- 43
{ jet , PTLep  cos  ( j , Lep)}
1942 +- 44
the observed number
1876 +- 28
Lumi 200pb-1
12
Introduction
Matrix Construction
Estimation
Conclusion
The estimated No. of btagged jets
Estimated No. of btagged jet of W+2j events
based on 3-dimensional eff matrix from W+3j events
Variables
Estimated No. of btagged jet
{PT jet , jet , PTLep  cos  ( j , Lep)}
1881 +- 44
1886 +- 45
1914 +- 47
1944 +- 46
1876 +- 28
{PT jet , jet , cos  ( j, ET )}
{PT jet , PTLep  cos  ( j, Lep), cos  ( j, ET )}
{ jet , PTLep  cos  ( j , Lep), cos  ( j , ET )}
The observed number
Lumi 200pb-1
13
From the No. of jet to No. of Events
Introduction
Matrix Construction
Estimation
Conclusion
• The No. of b-tagged jets is not as useful or handy as the
No. of b-tagged events.
• But if we select the events(2j) with
– one jet(   2.5 )
– the other one(   2.5)
• Then
N
jet ,| jet | 2.5
tag
N
event
tag
• Why that?
– t-channel selection (CSC)
• at least 1 b-tagged jet with Pt>50 GeV and |eta|<2.5
• highest Pt untagged jet |eta|>2.5
14
The estimated No. of btagged events
Introduction
Matrix Construction
Estimation
Conclusion
Estimated No. of btagged events of W+2j
based on 2-dimensional eff matrix from W+3j
Variables
Estimated btagged events
{PT jet , jet }
196 +- 5
262 +- 6
248 +- 6
249 +- 6
199 +- 5
198 +- 5
202 +- 10
{PTLep  cos  ( j , Lep)，
cos  ( j , ET )}
{PT jet , cos  ( j , ET )}
{PT jet , PTLep  cos  ( j , Lep)}
{ jet , cos  ( j , ET )}
{ jet , PTLep  cos  ( j , Lep)}
the observed number
Lumi 200pb-1
15
Introduction
Matrix Construction
Estimation
Conclusion
The estimated No. of btagged events
estimated No. of btagged events of W+2j
based on 3-dimensional eff matrix from W+3j
Variables
Estiamted btagged events
{PT jet , jet , PTLep  cos  ( j , Lep)}
194+- 5
194 +- 5
252 +- 7
199 +- 5
202 +- 10
{PT jet , jet , cos  ( j, ET )}
{PT jet , PTLep  cos  ( j, Lep), cos  ( j, ET )}
{ jet , PTLep  cos  ( j , Lep), cos  ( j , ET )}
The observed number
Lumi 200pb-1
16
Introduction
Matrix Construction
Estimation
Conclusion
Data-driven method
• Till now, all studies are on W+jets events.
• But with real data coming, all matrices are from
data.
W  jets
data
tt
 3 j ( x) 
Ntagged (x)
N
W  jets
total
( x)

Ntagged (x)  Ntagged (x)
N
data
total
( x)  N
tt
total
( x)
• where tt rate could be estimated for example from
4-jets events
• Then, apply the matrix of efficiency to 2j bin
W  jets
W  jets
Ntagged
(
x
)


(
x
)

N
,2 j
3j
total ,2 j
W  jets
data
tt
SingleTop
Ntotal

N

N

N
,2 j
total
total
total
• which is under study.
17
Introduction
Matrix Construction
Estimation
Conclusion
Conclusion
• Several variables are promising
PT , , cos 
jet
jet
jet , ET
Lep
T
,P
cos 
jet , Lep
• High dimensional matrix is more closer to the
truth (but need higher statistics)
• In the future, with real data, the binning over
variables will be studied in details.
• Data-driven method is currently under study
18
Backup
19
20
Unpromising variables
21
Unpromising variables
22
Unpromising variables
Rebinning on 3D Matrix
• Because of low statistics on 3-D histogram, low
number binning scheme is needed here:
PT
6 bins: 30.0,60.0,90.0,120.0,150.0,190.0,250.0GeV

cos  ( j, ET )
6 bins: -2.5,-1.67,-0.83,0.00,0.83,1.67,2.5
PTLep  cos  ( j, Lep)
6 bins: -1,+1 (uniform)
6 bins: -200.0,-100.0,-50.0,0.0,50.0,100.0,200.0GeV
23
1-D Matrix(Jet No.)
Variables
Estimated btagged jet
PT jet
1912 +- 44

1959 +- 43
jet
cos  ( jet , ET )
1997 +- 44
PTLep  cos  ( jet , lep )
1983 +- 241
The observed number
1876 +- 28
Lumi 200pb-1
24
1-D Matrix(Event No.)
Apply the matrix to the events with
only one jet(   2.5 ) and the other one(   2.5 )
Variables
PT
jet

cos  ( jet , ET )
jet
PTLep  cos  ( jet , lep )
The observed number
P
Estimated btagged
event
jet
T
250 +- 6
200 +- 5
264 +- 6
262 +- 93
202 +- 10
Lumi 200pb-1
25