Feature Studies

[kratsg]

Correlations between oJet.subjetsPt and the leading tJet.gTowers

• eta/phi distance differences
• energy correlations (before and after pileup subtraction?)

[kratsg]

Ideally we want 3 types of plots:

• 2D graph of efficiency (Y-axis) vs NgTower >= X (X-axis)
  • Overlay signal (ZH and ttbar) and background (dijets) on the same plot
  • Want 3 or 4 versions of this with different gTower ET selections
• 2D graph of signal efficiency (X-axis) vs. background rejection (Y-axis)
  • i.e. signal efficiency vs. inverse background
  • Overlay different NgTower selections
  • Overlay different signals (ZH and ttbar)
• 2D graph of signal efficiency (X-axis) vs. trigger rate (Y-axis)
  • Overlay signals (ZH and ttbar) ... background is already the same

[kratsg]

There is only going to be 5 points on the x-axis (0 gTowers, 1 gTower, 2 gTowers, 3 gTowers, 4 gTowers). Is that what you're expecting?

denominator: offline selection of Nsubjets == 2 [W] or == 3 [boosted top] numerator: denominator with added selection of N(gTowerEt > M GeV) > Z

where M is per plot (such as 15 GeV, or 20 GeV) and Z is the number of gTowers (1,2,3) \geq than a threshold M.

This is going to be interpreted as "given an offline selection of Nsubjets, how efficient is it to select Z gTowers > M GeV"

[kratsg]

Can you just make individual points for the efficiencies and not lines? Can you explain why you have 100% efficiency for a selection that requires at least 5 gTowers above 10 GeV? At some point this selection should be inefficient. Perhaps the 10 GeV cut is just low enough that we down take the hit when requiring 5 of them, but we would see an inefficiency if the cut were raised to 20 or 30 GeV. Can you remove the requirement that the gFEX jet exists from the denominator? We want to know the total efficiency. Can you add an offline jet pT cut of pT > 200 GeV and mass within 100-220 GeV? (For the denominator) Can you add a gFEX jet pT cut of 200 GeV? (for the numerator)