2018 CRs Nick Amin September 2, 2018
Overview (1) ⚫ Previously looked at 2017 CRs with an updated fake rate ⚫ Have ntupled the latest golden JSON of 2018 data (35.5fb -1 ), so repeat the CR plotting with 2018 comparing against 2017 MC ⚫ Using 2017 parameters for 2018 (JECs, e ff ective areas, ID/Iso, b-tag WP , scale factors, nmiss==0, fake rates, flip rates) except pile up reweighting is modified to scale 2017 MC to 2018 PU distribution ⚫ Isolated trigger strategy courtesy of Laurent’s slides here • Explicitly, the babymaker uses these • 2018 is essentially same as 2017, and I’ve checked with brilcalc the e ff ective luminosity for the ee,e 𝜈 , 𝜈𝜈 paths each match the golden json lumi ⚫ Some caveats • Small fraction of corrupt/tail jobs, which should total around half a %. Eventually I will include these. • No ISR reweighting for 2018 � 2
Overview (2) ⚫ Essentially copying the format of 2017 CR slides (from here), so they should be flippable ⚫ Several combinations of years • 2017 (35.5fb-1): http://uaf-10.t2.ucsd.edu/~namin/dump/plots/plots_crs2017_Sep2/ • 2018 (41.5fb-1): http://uaf-10.t2.ucsd.edu/~namin/dump/plots/plots_crs2018_Sep2/ • 2017+2018 (77.1fb-1): http://uaf-10.t2.ucsd.edu/~namin/dump/plots/plots_crs20172018_Sep2/ • 2016+2017+2018 (112.9fb-1): http://uaf-10.t2.ucsd.edu/~namin/dump/plots/plots_crsrun2_Sep2/ • 2018 compared to 2017 data only : http://uaf-10.t2.ucsd.edu/~namin/dump/plots/ plots_crcomparedata20172018_Sep2/ ⚫ Again, careful since there’s thousands of plots in total (~150MB) ⚫ CRs, search string is lowercase of below items • BR • OSLOOSE (Z-dominated) • OS (tt ̅ -dominated) • CRZ • CRW • HTNB1 (fake-enriched, data-driven) • HTNB1 (fake-enriched, MC truth-fakes from tt ̅ ) ⚫ I’ll go through the same CR plots as before, and then show some interesting features/disagreements between the years at the end � 3
OSLOOSE (Z-dominated) ⚫ Selections: p T >25/20, OS pair ⚫ Good overall agreement (especially for ee pairs, 𝜈𝜈 could get better with SFs) ⚫ MET looks good! ⚫ Also, since MC is is not scaled to data, Z-counting in bottom left plot shows trigger paths look OK � 4
OS (tt ̅ -dominated) ⚫ Selections: p T >25/20, OS pair, MET>50, Njets ≥ 2, H T >300 ⚫ Some disagreement in e 𝜈 normalization ⚫ A bit of a normalization discrepancy here � 5
HTNB1 (fake-dominated) ⚫ Selections: p T >25/20, SS pair, MET>50, Njets ≥ 2, Nbtags==1, no H T >300 ⚫ Note: because of new fake rate prescription, must require raw p T >18 for L!T leptons when applying fake rate, or else we get an over-estimate ⚫ Not bad, especially since many things need to be updated to 2018 still � 6
HTNB1MC (fake-dominated) ⚫ Selections: p T >25/20, SS pair, MET>50, Njets ≥ 2, Nbtags==1, no H T >300 ⚫ This slide has truth-matched MC fakes from tt ̅ (flip with previous slide) � 7
CRW ⚫ p T >25/20, SS pair, MET>50, 2 ≤ Njets ≤ 5, Nbtags==2, H T >300, Nleps==2 ⚫ Data/MC = 1.45 ± 0.16 (no syst) � 8
CRZ ⚫ p T >25/20/20, SS Z-veto, MET>50, Njets ≥ 2, Nbtags ≥ 2, H T >300 ⚫ Data/MC = 1.22 ± 0.26 (no syst) � 9
BR (1) ⚫ p T >25/20(/20), SS, MET>50, Njets ≥ 2, Nbtags ≥ 2, H T >300 ⚫ Data/MC = 1.56 ± 0.14 (no syst) � 10
BR (2) ⚫ p T >25/20(/20), SS, MET>50, Njets ≥ 2, Nbtags ≥ 2, H T >300 ⚫ More plots wrt previous slide ⚫ ~2 𝜏 excess for charge>0 • ttW needs to be scaled up? • Slide 8 shows ttW would need to be scaled up by a factor of ~2, which would essentially kill this excess • Wouldn’t charge information help constrain ttW more in the fit? � 11
Misc comparisons � 12
Misc checks (1) ⚫ Check electron 𝜚 for the fake-dominated CR • Fake rate maps are the same for 2017 and 2018, so we can get relative loose-not-tight yields here • HEM gap — 𝜃 in [-3,-1.3], 𝜚 in [-1.57,-0.87] • Doesn’t seem like there’s a noticeable deviation in that region 2017 2018 � 13
Misc discrepancies (1) ⚫ The Njet==4 bin is pretty high all years in CRW • For bottom right ("run2"), data/MC is 1.51±0.14 in that bin → if we only allow ttW to float, this would give a ttW SF of ~2.2 2017 2017 + 2018 2018 2016 + 2017 + 2018 � 14
Misc discrepancies (2) ⚫ Lepton 𝜚 for electrons in the Z-dominated OS region has very ine ffi cient spots in 2018 • Pixel issue isn’t worse, so is this an artifact of using the same ID/Iso WPs as in 2017? 2017 2018 � 15
Misc discrepancies (3) ⚫ As mentioned on slide 11, lepton charge in baseline region has large excess in positive bin. Compare it with 2017 explicitly… ⚫ Is the disagreement just stats? � 16
Summary ⚫ 2018 data similar to 2017 ⚫ Discrepancy with • electron 𝜚 between the two years • ttW normalization � 17
Backup � 18
Recommend
More recommend
