60 likes | 253 Views
p1. delta phi method. Yousuke Kataoka, Naoko Kanaya, Shoji Asai (University of Tokyo). p2. introduction. motivation Currently delta phi control region is used to determine global QCD factor We extend the method to estimate the shape ,
E N D
p1 delta phi method Yousuke Kataoka, Naoko Kanaya, Shoji Asai (University of Tokyo)
p2 introduction • motivation • Currently delta phi control region is used to determine global QCD factor • We extend the method to estimate the shape, • in other words, QCD numbers in any MET, Meff region • advantages • there is no extrapolation in MET, Meff • control region is exactly the same with signal region in MET, Meff • most systematics (JES, cut bias, etc) doesn’t propagate • almost pure data-driven and no assumption on the MET source, • so respect all detector, heavy flavor, shower leak, even unknown sources • simple, easy, no extra data (just keep the events dphi<0.4 from SR) • disadvantage • statistics error contributes (several events in 1TeV region) • due to large factor of dphi ratio, error is suppressed and gives reasonable estimation • difficult to understand(validate) dphi ratio at high MET, Meff region • we use conservative value or low MET value (asymptotic sharpened)
Brief outline of dphi method p3 Signal region (dphi>0.4) Control region (dphi<0.2) Side-band region (dphi=0.2~0.4) typical shape of min dphi (4jet, met>130) x R (QCD) step4 QCD Nside QCD Nsig Ncont step3 x R (non-QCD) step1 step2 Non-QCD non-QCD (SM,SUSY) in signal region except for min dphi cut step1. get number of non-QCD(SM,SUSY) events in side-band region (dphi=0.2~0.4) assume Nside (non-QCD) ~ Nside (data) step2. get number of non-QCD events in control region (dphi<0.2) by multiplying the ratio Ncont (non-QCD) = Nside (non-QCD) x R(non-QCD:cont/side) step3. get number of QCD events in control region Ncont (QCD) = Ncont (data) - Ncont (non-QCD) step4. get number of QCD events in signal region (dphi>0.4) by multiplying the ratio Nsig (QCD) = Ncont (QCD) x R(QCD:sig/cont) contamination of QCD not propagate so much Inputs Ncont, Nside from data R(non-QCD:cont/side) R(QCD:sig/cont)
p4 R(non-QCD:cont/side) = Ncont(non-QCD) / Nside(non-QCD) • non-QCD events have similar distribution among processes, even in SUSY signals • due to the fact that “real-MET not correlate jets” • R is almost determined by probability(not physics) soMC is relatively reliable • R is similar and we can treat non-QCD BG (and even SUSY) as a whole • futhermore, systematics related to R is suppress by factor R(non-QCD) x R(QCD) ~ O(10) R = N_s / N_c Ncont Nside R = 1.09 +- 10% number here is an example, we prepare the number for each signal region (see last slide)
p5 R(QCD:sig/cont) = Nsig(QCD) / Ncont(QCD) It’s difficult to understand(validate) especially high MET, Meff region(signal region) because no pure QCD distribution available in data and MC is statistically limited but we can use conservative value R in low MET region for data and MC distribution gets sharpened at higher MET because large MET is rigid againt fake MET smearing we can use low MET value as conservative MET = 130GeV ~ 150GeV (4jets) Ncont Nsig Ncont Nsig MC(QCD) R = 0.11+-0.05 <0.16 Data - nonSM(MC) R = 0.07 +0.05 (20% sys. of nonSM) conservative value for high MET region * 0.16 for 4jets, 0.08 for 3jets, 0.11 for 2jets pour stat.
Estimation for each signal regions @ 163pb-1 p6 Inputs for each signal regions • pour statistics in MC • use conservative • 0.16 for 4jets • 0.08 for 3jets • 0.11 for 2jets Xsec from MC Estimation for each signal regions