Update on ttbar background
Data in CR= 2mu+2j50+T6>0.05modelled in MC by sum of all SM backgrounds + SUSY with CR cuts
Background ETmiss shape in Signal Region b(ETmiss)
modelled in MC by sum of all SM backgrounds
b(ETmiss)
100 GeV
2mu+2j50background only
c(ETmiss)
2mu+2j50+T6>0.05a(ETmiss)=b(ET
miss) +s(ET
miss)
Data in 2mu+2j50 signal + backgroundcontains signal and backgroundmodelled in MC by sum of all SM backgrounds + SUSY
b(ETmiss) is shape for the
backgrounds(ET
miss) is shape for the signalin 2mu+2j50
Same shape of ETmiss b/w bkg CR and SR <=> b(ET
miss)=k c(ETmiss) Need
to find kAll events in mu2+2j50a(ET
miss)=b(ETmiss) + s(ET
miss) and k = b(ETmiss) / c(ET
miss) but we do not have b(ET
miss)
If s(ETmiss)<< b(ET
miss) in a certain region NR (normalisation region) then for ETmiss
in NR a(ET
miss) ≈b(ETmiss) and
k = a(ETmiss) / c(ET
miss) ≈ b(ETmiss) / c(ET
miss) in NR k ≈ ∫ b(ET
miss) dETmiss / ∫ c(ET
miss) dETmiss integration over NR
OBSERVABLE DISTRIBUTIONS a(ETmiss ) and f(ET
miss)
NOT DIRECTLY OBSERVABLE b(ETmiss)
b(ETmiss) / c(ET
miss) = k => verify by plotting total background divided by
ETmiss in CR
s(ETmiss) / b(ET
miss) << 1 in NR => verify by plotting s(ETmiss) / b(ET
miss)
What are the approximations being made?
Exact equations
a(ETmiss)=b(ET
miss) + s(ETmiss)
k (1+ε(ETmiss)) = b(ET
miss) / c(ETmiss)
Formula effectively used: k = ∫ a(ETmiss) dET
miss / ∫ c(ETmiss) dET
miss over NR k = ∫ (b(ET
miss) + s(ETmiss)) dET
miss / ∫ c(ETmiss) dET
miss o=> Presence of signal in the NR leads to overestimate of k
Nbkg,SR = ∫SR
b(ETmiss) dET
miss
= k ∫SR
c(ETmiss) (1+ε(ET
miss) dETmiss
= k ∫SR
c(ETmiss) (1+ε(ET
miss) dETmiss
= k ∫SR
c(ETmiss) dET
miss + k ∫SR
c(ETmiss) ε(ET
miss) dETmiss
≈ k ∫SR
c(ETmiss) dET
miss
Size of the approximation due to shape assumptionNbkg,SR - k ∫
SR c(ET
miss) dETmiss = k ∫
SR c(ET
miss) ε(ETmiss) dET
miss
2mu + 2j50 CR= 2mu + 2j50 + T6 > 0.05
NR = 2mu + 2j50 + 60<ETmiss<80< GeV
SR = 2mu + 2j50 + ETmiss > 100 GeV
10pb-1 7 TeV
b(ETmiss) / c(ET
miss) = k => verify by plotting total background divided
by ETmiss in CR
Control Sample
At least 1 solution
T6> 0.10 T6> 0.05
CR statistics 25.5 29.2 30.1
T6> 0.10
T6> 0.05
Background shape from CR/ True background shape
ETmiss spectrum from CR is harder than for real backgroundif there is indeed SUSY, due to SUSY contamination.
At least 1 solution
At least 1 solutionNO SUSY
if no SUSY b(ETmiss) / c(ET
miss) = k is a very good approximation
In presence of SUSY, ETmiss spectrum from CR is harder than for real backgroundif there is indeed SUSY, due to SUSY contamination.
b(ETmiss) = ET
miss for background only c(ET
miss) = ETmiss in control region
Problem with statistics in CR @ 10pb-1
Control Region: T6> 0.10
Expect less than 1 eventin the useful region with ET
miss>100 GeV Useless=> Either find a different approach or looser cuts.
Signal Region
Normalisation Region
SU4 fraction in NR
SU4
NR=60 < ETmiss < 80 GeV
Normalisation Region
Number of events 10pb-1
SU4 fraction
Comment
60 < ETmiss < 80
GeV2.0 32%
40 < ETmiss < 60
GeV2.7 22.5%
20 < ETmiss < 40
GeV8.3 4% Using Z->mumu number of events to
normalise ttbar ETmiss spectrum-> overestimate. M(mu,mu) mass cut?
Results Based on previous slides chooseNR=40 < ET
miss < 60 GeVCR= 2mu + 2j50 + T6 > 0.05
Results 10pb-1 Based on previous slides chooseNR=40 < ET
miss < 60 GeVCR= 2mu + 2j50 + T6 > 0.05
If SUSY contaminationof NR and CR lead to overestimateof background, poorer exclusion limit
If no SUSY => excellent bkg prediction
SU4
A quick look at supersoft cuts @ 7 TeV 10pb-1
Cut Signal Bkg EFF S/√(B) med[ZW] Nmax σ95(pb) FOM Opal
----------------------------------------------------------------------------------------------------------------------------------
mu2_2j50_met100 2.8094 1.5435 0.0059 2.2613 1.8459 3.9485 67.2 676.2
mu2_mz_1j30_1j20_met30 5.3055 12.8320 0.0111 1.4811 1.3934 8.2491 74.3 729.5
mu2_1j100_1j50_met100 2.5451 1.0672 0.0053 2.4637 1.9284 4.0681 76.4 695.1
mu2_mz_2j20_met30 5.3545 13.0703 0.0112 1.4811 1.3941 8.8498 79.0 728.6
mu2_2j20_met30 5.7167 18.8571 0.0120 1.3165 1.2572 9.7251 81.3 802.5
mu2_1j30_1j20_met30 5.6677 18.3854 0.0119 1.3218 1.2614 9.8994 83.5 800.3
mu2_3j50_met100 1.5858 0.5554 0.0033 2.1279 1.6147 3.0093 90.7 1019.6
mu2_1j100_2j50_met100 1.4781 0.4610 0.0031 2.1770 1.6171 3.0062 97.2 1073.8
med[ZW] : Discovery significance G.Cowan, E. Gross, Stat. Forum December 2nd 2009 Eq.23
better approximation than S/√(B)
Nmax : median 95% CL upper limit on the number of signal events in SR
σ95 : 95% CL upper limit on signal cross section given "Asimov dataset"
FOM Opal : approximate FOM related to maximum possible number of signal events in SR
follows Nmax
Still no systematics included
mu2_1j30_1j20_met30
At least 1 ttbar solution
Normalisation region 2mu 1j30 1j20 20<MET<40 - should be 20-30 GeV ...
Composition of normalisation region
------------------------------------------------------------------Nevents in background region (MC) = 2.76907Nevents in background region (from CR prediction) = 3.2448Nevents in control region = 129.301Nevents in normalisation region = 26.496------------------------------------------------------------------
------------------------------------------------------------------Nevents in background region (MC) = 1.54353Nevents in background region (from CR prediction) = 0.519253Nevents in control region = 25.6899Nevents in normalisation region = 2.69708------------------------------------------------------------------
Soft Cuts (mu2_1j30_1j20_met30) mu2_2j50_met100
Outlook
Complete ttbar bkg with supersoft cuts and mZ cuts
Include systematics for ttbar background in FOM
Include other ballpark systematics in FOM Conclude on ttbar background at 10pb-1
Conclude on cuts for 10pb-1 analysis
Discussion ongoing concerning a softer SUSY model