First Attempts ABCD Method for background estimation of

First Attempts: ABCD Method for background estimation of W→eν Shu Li, Yanwen Liu Together with Elisabeth Petit, Fabrice Hubaut, Pascal Pralavorio Center of Particle Physics and Technology, USTC &Center de Physique des Particules de Marseille, CNRS-IN 2 P 3 / Université de la Méditerranée - Aix Marseille II 1

Datasets in use n 7 Te. V Collision data used: (~14 M evts) ü L 1 Calo. EM ESD: all of Run. Number 152166 -153599 ü L 1 Calo ESD: all of Run. Number 154810 -155160 n MC Samples used: ü mc 09_7 Te. V. 105802. JF 17_pythia_jet_filter. merge. AOD. e 505 _s 765_s 767_r 1302_r 1306 (10 M evts, serves as background with signal components excluded) ü mc 09_7 Te. V. 105805. filtered_minbias 6. merge. AOD. e 530_s 7 65_s 767_r 1302_r 1306 (7 M evts, signal sample) 2

Basic Selection Criteria in use based on WZObservation fundamental requirements on Twiki: https: //twiki. cern. ch/twiki/bin/view/Atlas. Protected/WZObservation 7 Te. V ü Pre. Selections: Ø Event Level: Ask for stable beam Lumi. Block based on Interesting. Runs Twiki: https: //twiki. cern. ch/twiki/bin/view/Atlas. Protected/Interesting. Runs 2010 GRL used: (going to update with new egamma GRLs) • data 10_7 Te. V. period. A. 152166 -153200_LBSUMM_Det. Status-v 02 repro 04 -00_eg_standard_7 Te. V. xml • data 10_7 Te. V. period. B. 153565 -155160_LBSUMM_Det. Status-v 02 repro 04 -00_eg_standard_7 Te. V. xml Good Lumi. Block with Stable beams required for all runs L 1_EM 2 trigger passed Good Primary Vertex with ≥ 3 tracks MET_cleaning implemented using Jet. Calo. Quality Tool (https: //twiki. cern. ch/twiki/bin/view/Sandbox/Jet. Calo. Quality) 3

üFurther Pre. Selections: Ø Object Level: Fiducial Cuts applied using Check. OQ Macro(Version 4, not the latest) to exclude dead OTX related clusters (Still updating) (/afs/cern. ch/atlas/groups/EGamma/OQMaps/OLD: check. OQ_v 4. C & Object. Quality. Maps_run 152166. root & Object. Quality. Maps_run 155118. root) ETclus>15 Ge. V |ηclus|<2. 0 without crack region (only within TRT coverage for e. Probability. HT, should be 2. 5 for Calo Isolation study) Author. Electron (author==1 || author==3 for MC AOD, author==1 for data ESD) is. EM Loose/Medium/Tight applied respectively Track quality requirements: n. SCTHits>=6, n. Pixel. Hits>0, n. TRTHits>=10 Basic shower shapes requirements: f 1>0. 01; weta 1, d. Emin. S 1, d. Emax. S 1, Wstot. S 1 >0. 4

ü ABCD Method 1. (data-driven only, need to verify in MC) Assume ABCD regions have similar ratio over each other for those background components of each region in the data: bgd. A/bgd. B=bgd. D/bgd. C So as to extract the signal component in “D” while removing the estimated background in it. (Should be verified based on MC) ü ABCD Method 2. (using MC information, limited due to MC and Data disagreement and statistics) Assume ABCD regions have similar ratio over each other for both signal and bgd between data and Monte Carlo: b 1 i+b 2 i+si=Ni, i=A, B, C, D b 1 i: 1 st kind of background (JF 17), b 2 i: 2 nd kind of background (filtered_Min. Bias) Ni: Expected from data si/si, b 1 i/b 1 j, b 2 i/b 2 j in data should be consistent with MC 3 equations and 3 unknown variables 5 Can be implemented with more or fewer background samples

Electron Probability for TRT High Threshold(variable named e. Probability. HT) Vs MET_Topo e. Probability. HT Vs MET_Topo • A: MET_Topo <= 25 Ge. V, e. Probability. HT >= 0. 9 • B: MET_Topo <= 25 Ge. V, e. Probability. HT < 0. 9 • C : MET_Topo > 25 Ge. V, e. Probability. HT <= 0. 9 • Signal Region D: MET_Topo > 25 Ge. V, e. Probability. HT > 0. 9 6

e. Probability. HT distribution after pre. Cuts with is. EM medium applied (Not Normalized) 7

e. Probability. HT Vs MET_Topo after pre. Cuts with is. EM medium applied (Not Normalized) 8

is. EM Loose (Not Normalized) A B C D Wenu (7 M) 305964 57413 254392 1332229 JF 17 (10 M) 4323 7347 33 92 Data (~14 M) 95 Sample /data Region A/B=D/C doesn’t agree 163 1 Very tight 9 20

is. EM Medium (Not Normalized) A B C D Wenu (7 M) 299627 56346 250544 1309123 JF 17 (10 M) 2268 1721 17 80 Data (~14 M) 54 Sample /data Region A/B=D/C doesn’t agree 35 1 Very tight 10 19

Is. EM Tight is not appropriate to be applied due to its correlation with the variable e. Probability. HT! 11

Calorimeter Isolation(Calo. Iso) Vs MET_Topo Calo. Iso Vs MET_Topo • A: MET_Topo <= 25 Ge. V, Calo. Iso <= 0. 1 • B: MET_Topo <= 25 Ge. V, Calo. Iso > 0. 1 • C : MET_Topo > 25 Ge. V, Calo. Iso >= 0. 1 • Signal Region D: MET_Topo > 25 Ge. V, Calo. Iso < 0. 1 12

Calorimeter Isolation distribution after pre. Cuts with is. EM medium applied (Not Normalized) 13

Calorimeter Isolation Vs MET_Topo after pre. Cuts with is. EM medium applied (Not Normalized) 14

is. EM Loose (Not Normalized) A B C D Wenu (7 M) 318890 44487 145977 1440644 JF 17 (10 M) 2108 9562 54 71 Data (~14 M) 55 Sample /data Region A/B=D/C doesn’t agree 208 15 4 17

is. EM Medium (Not Normalized) A B C D Wenu (7 M) 312709 43264 142109 1417558 JF 17 (10 M) 914 3075 30 67 Data (~14 M) 30 Sample /data Region A/B=D/C doesn’t agree 59 16 4 16

is. EM Tight (Not Normalized) A B C D Wenu (7 M) 312709 43264 142109 1417558 JF 17 (10 M) 144 11 11 144 Data (~14 M) 30 Sample /data Region A/B=D/C agrees with the assumption extremely well 59 17 4 16

Some possibilities for next step TRT High Threshold Probability still need to be further investigated since the results did not agree with assumption in method 1 (try with method 2? ) Try with New Recommended Robust is. EM Tight Cuts Update with new fundamental selection criteria(new GRL, OQMap, etc. ) with more new collision runs Some other way: 2 -D Matrix method (very simple way to carry on by using only one variable but may result in some disagreement between different variables) Change some values of the current cut criteria and evaluate the systematics 18