umd-lhcb / misid-unfold

Unfolding Muon misID samples
BSD 2-Clause "Simplified" License
1 stars 1 forks source link

Prepare misID weights for D* comb bkg #7

Closed yipengsun closed 2 years ago

yipengsun commented 2 years ago

Currently we only have the weights for B comb. bkg. as the unfolding weights are different between nominal and B USB region.

For D*, we need to have different weights for events IN and OUTSIDE the deltaM = (mD* -mD0) window only.

So we need 2 species, let's call them dst_nominal and dst_out. The selections are:

@Svende @manuelfs Do you think the cuts listed above are the cuts for the D* comb. bkg? If so, I'll start processing them and the ntuples containing the D* misID weights should be ready by next Mon.

yipengsun commented 2 years ago

Or, should we forget about the mass window cut on the B0 entirely? Because that cuts would highly favor events within the deltaM window.

Svende commented 2 years ago

Yes I think we don't need the B mass window cut but apply the D0 mass window cut for those weights. So essentially we need what you call dst_nominal & dst_out selections.

yipengsun commented 2 years ago

In the end, we'll do the following:

yipengsun commented 2 years ago

OK, I started the misID ntuple generation process. Should be done before tomorrow afternoon.

yipengsun commented 2 years ago

Latest misID ntuples added at rdx-run2-analysis:

ntuples/0.9.6-2016_production/Dst_D0-mu_misid

Use git-annex to download these ntuples.

yipengsun commented 2 years ago

Testing if we can apply b_m_ok && in_fit_range for D* samples.

Currently, we only require d0_m_ok && dst_m_ok (comb_d) and d0_m_ok && !dst_m_ok (comb_dsb). For the D* right sign sample, we have:

For the wrong-sign Pi sample (I believe for this samples, more events will FAIL the dst_m_ok cut):

Now if we require b_m_ok && in_fit_range in addition, for the same D* right sign sample:

For the wrong-sign Pi samples:

To conclude, I believe it is fine to apply the additional b_m_ok && in_fit_range cuts.

FYI @manuelfs @Svende

Svende commented 2 years ago

Thanks for looking into this @yipengsun! Following your conclusion that those cuts won't kill too many events, are you then planning to do another round of unfolding misID tuples adding those cuts?

yipengsun commented 2 years ago

I plan to add those after we finish the preliminary fit for both comb D* and comb B. Because:

  1. These cuts don't make a huge difference, so current weights are usable
  2. There might be problems/bugs, but we can't tell until we do fit at least once. I'd like to use this opportunity to fix them together.
yipengsun commented 2 years ago

We've discussed to split the reweighting in 4 regions. For one of the region (D SB B SB), some of the tagged species contain very few events (namely, the electron sample contains only ~30 events).

yipengsun commented 2 years ago

Looking at the unfolding result in the D SB B SB region: (top: before, bot: after):

Dst_dsb_bsb-tag_p Dst_dsb_bsb-true_p

I think the unfolding is still working, in the sense that we still recovered some true particles from the tagged ghost samples.

So, I'll proceed with the current split, which has 4 separate unfolding regions.