Maintenance script: Created page with " == [TopNewPhysics] Roadmap for data analysis == PageOutline === 0) Preliminary work to be done before data-taking === *framework and tools ready (see slides ..."

2015-08-26T12:29:02Z

Created page with " == [TopNewPhysics] Roadmap for data analysis == PageOutline === 0) Preliminary work to be done before data-taking === *framework and tools ready (see slides ..."

New page

== [TopNewPhysics] Roadmap for data analysis ==

[[PageOutline]]

=== 0) Preliminary work to be done before data-taking ===

*framework and tools ready (see slides ...)
*TopTreeAnalysis package should be cleaned/documented/reworked
*Analysis should be done @ 7 TeV to estimate the #events expected for a given luminosity and selection
*Instore a kind of weekly meeting for everybody (Michael&Stijn included) to exchange information (short report of meetings, hypernews ...)
*TopTreeAnalysis package has to be extended according to the test we want to do on data
- 1)b) - 2)c)
*Need tests to be performed on 7 TeV samples
- 1)c) - 2)c)

In parallel, the analysis for the inclusive search for new physics has to be developed on MC (study of observables, gof ...)

=== 1) With the first few data (~ 1 pb)===

-> Major goals:
a) testing that our tools (from RECO to TopTree + TopTreeAnalysis) are working 
b) "Validate" the reconstructed objects from data 
c) checking the validity of our hypothese for the ABCD method 

==== a) testing that our tools (from RECO to TopTree + TopTreeAnalysis) are working ====
*From day one, run over the available data and check if we retrive outputs. Check for failures, inefficiencies, inconsistencies ...
*Monitor the efficiency to retrieve a final TopTree from available dataset
*Make sure that all the tools are working well, modifications/developments could be made in case of missing items

==== b) "Validate" the reconstructed objects from data ====
[This section needs to have tools developed in advance]

*From the events available in the SD, we want to have the breakdown of all selection cut apply (histos, efficiency plots, tables)
*More generally check the stability of all cuts over the runs

- Focus on isolated muons in jetty environment (involve Michael)
*trigger rate ? (can't be compute with SD ... but from PD to go to SD)
*study of the quality cuts apply (from trigger muon to selected muon)
*compare RelIso var (data vs MC), breakdown per subvariables (TrackIso, ECALIso, HCALIso), per #jets (histos, efficiency plots, table)
*check muon ecal deposit, DR(m,j) cut (histo, efficiencies)
*d0Sign distribution, efficiencies as function of the variable,

- Focus on jet multiplicity (involve Stijn)
*study of the jet-id cuts
*eta-phi distribution of jets after Pt selection (look for hot cells)
*jet multiplicity after 2-3 different selections of jets

==== c) checking the validity of our hypothese for the ABCD method ====
[need test performed on MC first]
To test our methods, we need enough statistics in data to be able to distinguish between statistical fluctuations and possible bias.
To enrich our signal region:
*Lower Pt cut on jets 30-40 ↘ 20-25 GeV
*if it's not enough, maybe lowering the #jets required 4 ↘ 3
*relax a bit RelIso cut

- Test of the ABCD method
[need test performed on MC first]
*As there is not too much statisc in the SR, RelIso cut cout be relaxed at the beginning
*Check the correlation between RelIso & doSign
*Put a MET cut (MET<~20-30 GeV) to remove EWK-top events in the SR - see the impact on RelIso & d0Sign variable - check if prediction is equal to what we observed on data (true if no EWK-top contribution)
*Check stability of the prediction while changing the range of the regions
*Divide BD regions in 4 and check the consistency (ABCD method - closure test)
*While removing this cut, check that the estimated difference is compatible with EWK-top expectations

- Test the QCD shape estimation
[need test performed on MC first]
*Check correlation between RelIso and the observable we considerer (focus on only few distributions - to be defined)
*Divide the CR in 2 or more bins depending on the nof events (~1000 per histo) and compare the template obtained (momenta, Chi2, Komogorov-Smirnow)
*Compare those templates to the ones predicted to the MC, are we far from it ?
*If there is already enough events in the SR (low pt cuts, relaxed RelIso), compare estimation with data
*If the method fails, find a extrapolation procedure

Increase the selection criteria when luminosity increase

=== 2) With a luminosity of ~ 10-50 pb ===

-> Major goal: participate to a cross-section measurement
a) study of b-tagging 
b) study V+j estimation 
c) participate to the measurement 

==== a) study of b-tagging ====
*follow the alignment of the tracker
*follow the results obtain in the b-tag group
*focuss on the more robust b-tag algo and make distributions with data (compare to MC) #b-jets (as function of #jets & pt/eta(jet) )

==== b) study of W+j estimation ====
[need test performed on MC first]
[This section needs to have tools developed in advance]

In order to increase the statistic
*From 3 categories of #jets: 4-5->=6 go to 1 categorie: >=4
*To increase 3 b-jets categorie: use high effiency WP (“loose”)

In order to reject top events:
*Cut on a Chi2 (jet combination) and/or HT (Nttbar ~ 0)
-> Make Chi2 distribution (and reco-masses), compare to MC - Efficiency as function of the cut - find a safe cut

Test of the method
*Nttbar should be ~ 0 after Chi2 rejection (safe/high cut to not depend on JES)
*Check stability of the prediction while changing b-tagging algo & WP (maybe not too much freedom ... loose - semi-loose WP for trackCounting algo)
*Eb &Eudsc could be compare to MC or to estimation (b-tag group after MC-correction factor)
*Possibility to probe Vbb content (if there is no ttbar contamination). Compare to MC, gave a k-factor which could be apply after lowering Chi2-cut
*Control the Nttbar estimation while lowering the Chi2 (#Ntt as function of Chi2 cut)

The shape of W+j could be obtain with "int" variable like charge of lepton
*look charge of lepton. Template for V+j should be assymetric and the one for tt+j should'nt
*compare the template obtain to the MC-prediction (charge assymetry) - ~ closure test (should give #N(W+j) )

==== c) cross-section measurement ====
*keep in contact with other group working on the l+j channel
*follow the work on event selection
*give our bkg estimation results
*or plugin external results (trigger efficiency, muon selection efficiency ...) to obtain the cross-section

=== 3) With a luminosity of ~ 100 pb ===

-> Major goal: look for differential distributions
a) study the shape estimation methods 
b) obtain the first differential distributions 
c) perform a goodness-of-fit test distribution per distribution -> MC tuning ?! 
d) combine observable in a goodness-of-fit test 

==== a) study the shape estimation methods ====
*extension of the previous work for QCD
*estimation of W+j
*estimation of tt+j

.. for next ... we still have time ...

{{TracNotice|{{PAGENAME}}}}

RoadmapDataAnaTopNewPhysics - Revision history

Maintenance script: Created page with " == [TopNewPhysics] Roadmap for data analysis == PageOutline === 0) Preliminary work to be done before data-taking === *framework and tools ready (see slides ..."