Status Report for Standard Model WH(bb) Analysis Yanhui Ma VH(bb) Mee)ng @LAL 2016-11- 16
Introduction Ø Search for a SM Higgs boson produced in associa?on with a vector boson(w/z), and decaying to a pair of b- quarks. Ø 3 channels Ø Not yet observed in Run1 l ATLAS: 1.4 (2.6) σ in observed (exp.) l CMS VH: 2.1 (2.1) σ l ATLAS + CMS: 2.6 (3.7) σ Ø Will focus on 1 lepton channel in this talk 2
Data and Simulated Samples Ø Total 13.2 fb-1 data was used for the ICHEP analysis. l 3.2(2015) + 10(2016) 3
Object Selec?on Ø Overlap removal algorithm l The electron will be removed when: 1. A reconstructed muon and electron share the same matched ID track 3. Any electrons reconstructed within R < 0.4 around the jet axis of a surviving jet l The Muon will be removed when: 5. A muon is reconstructed within a cone of radius R = 0.4 around the jet axis of any surviving jets l The Jet will be removed when: 2. The closest jet to an electron within a cone of radius R = 0.2 around the electron 4. A jet is reconstructed within R < 0.2 around a muon and the jet has fewer than three associated tracks Ø B-tagging 6. Jets that are reconstructed within a cone of radius R = 0.2 around the axis of a tau candidate l MV2c10 l 70 % WP l truth tagging is used for V+cc, cl, light 4
B- jet energy correc?on Ø It is important to find muons derived from B- hadron semileptonic decays for bemer b- jet energy resolu?on Ø in about 12% of b- jets, there is an muon and a neutrino within the b- jet cone. Ø The choice for ICHEP is :PtRecollbbOneMuTruthWZNoneNewTrue Ø Muons inside jets are required to have pt > 4 GeV, η < 2.7 and medium quality Ø Choose the closest muon in R to the jet. The muon 4- vector is added to the jet 4- vector, while the 4- vector of the energy deposited by the muon in the calorimeter is removed. 5
Event Categoriza?on and Selec?on Ø The 1- tag region is not currently included in order to simplify the fit model and systema)c uncertain)es. Ø In the 1- lepton channel, due to the modelling difficul?es related to the es)mate of the high mul)jet background contamina)on, the [0, 150] GeV region is not considered. 6
Event Categoriza?on and Selec?on Ø Muon channel are triggered using MET trigger à to increase the trigger efficiency for single high pt muons Ø The combina)on of the MET trigger and muon trigger can offer only a small increase in the signal acceptance à do not used to simplify the analysis Ø The trigger efficiency, for events with a reconstructed MET = 150 GeV, is about 85%,the efficiency plateau is reached at about MET = 180 GeV. Ø In order to handle properly triggers in simula)on, we use the random run number proivded by pile- up reweigh?ng tool to define if the event is associated to 2015 or 2016 data. 7
Event Categoriza?on and Selec?on Ø Due to the increased likelihood for a jet to be reconstructed incorrectly as an electron compared to a muon, separate an?- mul?jet selec)ons have been op)mised for the electron and muon channels. Combined Electron channel Ø Obviously overshoot for data can be seen for MET distribution in low MET bins Muon channel Ø mostly due to the electron channel, muon channel looks fine 8
Event Categoriza?on and Selec?on Combined Electron channel Muon channel 9
Event Categoriza?on and Selec?on Ø In the 1- lepton channel, the following addi)onal selec)on criteria are applied for the dijet mass analysis cross- check : l R(jet1, jet2) < 1.8 (150 <= ptv < 200 GeV), <1.2 (200 <= pvt GeV); l mt (W) < 120 GeV 10
Mul?- jet in 1 lepton channel- - - fake factor method Ø Dijet CR: exactly 1 signal jet fake factor f = CR1/CR2 SR = f x CR3 Ø The electroweak contamina)on is removed from the di- jet control regions using the simulated MC predic?on Ø the normaliza)on of the electroweak processes in the di- jet control regions is scaled with fits in high lepton purity regions: 150 < MET< 250 GeV. Ø Fake factors derived for 5 pt(μ) x 3 η(μ) regions using di- jet control regions Ø MET in fake electron events is strongly correlated with the energy mis- measurement of the fake electrons 11
Mul?- jet in 1 lepton channel- - - fake factor method 12
Mul?- Variate Analysis Ø based on Run1 experience Ø no change on configura)on Ø pile- up reweigh)ng is not applied, truth tagging(forv+cc,cl,l sample) is applied, to increase the MC sta)s)cs to get a more stable, op)mal training. 13
Mul?- Variate Analysis Ø adding mtop and ΔY(W, H) for 1 lepton channel(expected increase of significance ~ 7 %) 14
Mul?- Variate Analysis Ø 1 lepton channel BDT output for 2tag2jets(left) region and 2tag3jets(right) region. Ø Expected significance comparison of BDT- based and cut- based analyses 15
Mul?- Variate Analysis Ø 1 lepton channel BDT output for 2tag2jets(left) region and 2tag3jets(right) region. Ø Expected significance comparison of BDT- based and cut- based analyses 16
Pre- fit data- MC comparison- - - - SR- - - - 2tag2jet You can find all the plots in(thanks Charles for providing the script to create the web page): htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 2tag2jet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 2tag2jet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 2tag2jet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 2tag2jet htps://yama.web.cern.ch/yama/plots_15thnov/ Events (Data-Bkg)/Bkg 1000 800 600 400 200 0 0.3 0.2 0.1 0.3 0.2 0.1 2.5 2.6 2.7 2.8 2.9 3 3.1 (V,b,b) ATLAS Internal 2 KS -1 Stat 1.04 0.719 Ldt = 13.18 fb s = 13 TeV Syst 0.642 0.554 V WH l bb, 2 tag, 2 jets,p >= 150 GeV T Shape 0.642 0.554 Data VH 125 *50 diboson multijet Z+bb Z+bc Z+bl Z+cc Z+cl W+bb W+bc W+bl W+cc W+cl W+l single top ttbar (Data-Bkg)/Bkg Stat Stat+Shape Stat+Sys
Events Yields - - - - SR- - - - 2tag2jet Ø Lew : My results Ø Right: ICHEP results 22
Pre- fit data- MC comparison- - - - SR- - - - 2tag3jet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 2tag3jet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 2tag3jet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 2tag3jet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 2tag3jet htps://yama.web.cern.ch/yama/plots_15thnov/
Events Yields - - - - SR- - - - 2tag3jet Ø Lew : My results Ø Right: ICHEP results
Pre- fit data- MC comparison- - - - SR- - - - 1tag2Pjet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 1tag2Pjet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 1tag2Pjet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 1tag2Pjet htps://yama.web.cern.ch/yama/plots_15thnov/
Pre- fit data- MC comparison- - - - SR- - - - 1tag2Pjet htps://yama.web.cern.ch/yama/plots_15thnov/
VH Fit : ICHEP Results Ø Significance : 0.42 (1.94) obs. (exp.) l 0- lepton 1.02 (1.45) l 1- lepton 0.10 (1.04) l 2- lepton - 0.17 (1.14) 34
Back Up 35
Sta?s?cal Treatment Ø Binned maximum likelihood fit simultaneously in 8 region Ø Input variable : BDT Ø Parameter of interest : Common signal strength μ= σ fit / σsm across channels and regions Ø Associa)ng either floa)ng scale factors (Zbb, Wbb, tbar) or nuisance parameters (NP) for the normalisa)on Ø Shape systema)c uncertain)es: mbb, ptv Ø VZ Fit valida)on (BDTvz)à Unblind VH Fit l Re- training BDT to extract VZ signal l setup as much as possible consistent with main fit l Significance : 2.97 (3.18) obs. (exp.) 36
Experimental Systema?c Uncertain?es Ø Luminosity Ø Pileup Ø B- tagging SysFT EFF extrapola)on SysFT EFF extrapola)on from charm SysFT EFF Eigen Light 0/1/2/3/4 SysFT EFF Eigen B 0/1/2/3/4 SysFT EFF Eigen C 0/1/2/3/4 Ø Jets SysJET JER SINGLE NP JES 19 NP Ø Muons MUONS ID/MS/SCALE MUON ISO SYS/STAT MUON EFF SYS/STAT/ TrigStatUncertainty/TrigSystUncertainty MUON TTVA SYS/STAT Ø MET SysMETTrigStat SysMETTrigTop SysMETTrigZ SysMET JetTrk Scale SysMET SowTrk ResoPerp SysMET SowTrk ResoPara SysMET SowTrk Scale Ø Electrons EG RESOLUTION ALL EG SCALE ALL EG EFF ID/Iso/Reco/TriggerTotalCorr Uncertainty Ø MC stat 37
VH Fit : Valida?on : NP Pulls Ø Jet, b- tag, MET, lepton Ø Asimov (red), data (black) Ø Overall good behavior 38
VH Fit : Valida?on : NP Pulls Ø Zjets, Wjets, Top, Diboson Ø Asimov (red), data (black) Ø Overall good behavior 39
VH Fit : Valida?on : Correla?on Matrices Ø Asimov (lew) and data (right) show the same patern 40
VH Fit : Results : Posfit plots 41
VH Fit : Results : Posfit plots 42
VH Fit : Results Ø Significance : 0.42 (1.94) obs. (exp.) l 0- lepton 1.02 (1.45) l 1- lepton 0.10 (1.04) l 2- lepton - 0.17 (1.14) 43