Differences between data and MC cut

[yipengsun]

Here I'll document the differences between data and MC cuts and my implementations. Some of the cuts requires #83 to be finished first.

The general idea is: We DONT require truth-matching, except for Mu (because we have a dedicated Mu mis-ID samples to study the effects of misID).

cut name	data	MC
K PID	PIDK > 4 & !isMuon	apply PID weights
Pi PID	PIDK < 2 & !isMuon	apply PID weights
Mu PID	isMuon & PIDmu > 2 & PIDe < 1 & BDTmu > 0.25	has a true Mu and apply official PID weights and uBDT weights (The true Mu requirement is enforced at truth-matching already)

[yipengsun]

I separated the UBDT cuts from the regular PID cuts, because candidates that fail these cuts are still used as some control samples.

[yipengsun]

For run 2, 95% candidates that pass the regular selection cuts also pass the UBDT cut. This is true for both D0 and D* trees.

[yipengsun]

Phoebe has:

• The DD skim cut regarding iso_NNkw at: https://gitlab.cern.ch/bhamilto/rdvsrdst-histfactory/-/blob/master/proc/redoHistos_Dst.C#L1584
• Which comes into play for 1OS/D** at: https://gitlab.cern.ch/bhamilto/rdvsrdst-histfactory/-/blob/master/proc/redoHistos_Dst.C#L1642
• For 2OS, https://gitlab.cern.ch/bhamilto/rdvsrdst-histfactory/-/blob/master/proc/redoHistos_Dst.C#L1618

I don't see truth-matching there.

[yipengsun]

We confirmed with Pheobe that for K, Pi, we don't require the particle to be truth-matched. This is because we don't have a separate misID sample.

[yipengsun]

Here I document the main difference between the data and MC cuts, as shown in our run 2 postprocessing YAML:

• MC has NO PID cut (note how d0_pid_ok and mu_pid_ok are both set to true)
  • The mu_ubdt_ok is unused for MC, we'll apply it as a weight (for now the wpid_ubdt is set to 1.0)
• We require MC events to pass Phoebe/Alex's truth matching
• We require MC to have a valid HAMMER weight (except for DDX and D**s samples)
  • Note that before feeding input to HAMMER, I have a naive truth-matching which should be looser than Phoebe/Alex's
  • We truth-matching the Muon ONLY.
  • Then we require that HAMMER processed the event fine
• The L0 triggers are applied as weights, so the l0 boolean is set to true for MC.
  • The emulated L0 triggers are applied as a weight wtrg. Note that we require either D0 L0Hadron TOS or B L0Global TIS, so the L0 weight is defined as :d0_l0_hadron_tos_emu+wtrg_l0_tis - d0_l0_hadron_tos_emu*wtrg_l0_tis
• Hlt1 is applied normally since they are emulated as booleans.
• There's no Hlt2 cut since it's applied in DaVinci. So hlt2 is set to true for MC as well.
• The skim cuts are applied as weights, and the scheme is consistent w/ what we discussed at here (private)
• This commit has a global weight cut (because in one of the samples there's a single candidate that has a w of ~120, while the rest is below 10). Still, I think it's a bad idea to cut on weight so early so I'll remove it before committing the step-2 MC ntupes

FYI @manuelfs @Svende @afernez

[manuelfs]

This is very helpful, thank you very much Yipeng 🙏