Color code for answers:

- Comment is acknowledged and answered

- Authors are working on answering the comment

- Comment requires further work to be addressed or need attention from the internal reviewer regarding a specific issue

- We do not agree with the comment and arguments are given

AN-v3

Comments from Ankita

• Please use the recommended luminosity value consistently.
  • We are using the recommended values for the luminosity as well as their corresponding uncertainties. We use the values from:
    https://twiki.cern.ch/twiki/bin/view/CMS/LumiRecommendationsRun2
    
• Section2: Would be good to mention how you remove overlap b/w different datasets.
  • The text will be updated to reflect that statement and will be available in the v4 of the AN.
• L88: preVFP/post VFP (typo).
  • Done
• L95: do we include leptons from the tau decays as the signal?
  • We do not include the leptons from the the tau decays in the signal nor in the control region.
• Trigger eff. scale factors: just to avoid confusion perhaps better to drop the default value I.e bins with no events should be kept empty.
  • These values are scale factors and they represent simply the default value used when no scale factor is applied which is 1.0 in this case. We will add this statement in the Figure description to avoid confusion.
• L212: The collection of PF candidates after the Charged Gadron Sunbtraction is used, wanna run a spell check :wink:
  • Touché! Done
• L224: JECS for 2016, is this the latest greatest tag? I found `Summer19UL16APV_V7_MC` and `Summer19UL16 _V7_MC`.
  • We are using `Summer19UL16APV_V7_MC` and `Summer19UL16 _V7_MC` and these are the latest available JECs for the Run-2 Utra-Legacy samples. The text will be updated with that information in v4.
• L234: which tagger are you using? Table 5 suggest [DeepCSV] and here it says [DeepJet]. Please clarify?
  • We are using the `DeepJet` across the whole analysis. The text will be updated to remove the confusion.
• Section 4.2, For this EWK+QCD interference why do you omit taus? I see you use a factor 3 to account for missing lepton flavours but this does not take into account tau branching fractions. The effect might be small but I’m curious why not consider all flavours to begin with?
  • When computing the interference contribution in Magraph, you cannot require Z bosons in the intermediate state, so you must look at the inclusive production of a particular flavor of leptons. We did this for muons and multiplied by three. For the purposes of the interference calculation, tau's are treated as stable final-state particles, so it is not necessary to consider how they decay.
• L332: Could you add some distributions to show different components, EWK, EWK+QCD and interference. How many events in total you have from the interference to begin with?
  • We have done an estimation in the past that gives cross section ~0.0003838 pb for interference. VBS sec is 0.00308. Number of events from interference: 4.8 events before MJJ and EtaJJ cuts, 1.0 events after. We plan to implement this as a relative uncertainty on the VBS signal: (30+-4)%*(1/4.8)*(EW-only after EtaJJ and Mjj cuts)/(EW-only before EtaJJ and Mjj cuts). This will be done in the next iteration of the analysis.
• Section6: Add distributions for the input variables. Also, It becomes clear later on that you use GNN but would be good to add some concluding remarks here why you prefer GNN over the MLP which should be after figure 22(?). Actually, also the loss plots too.
  • We have made a comparaison between the MLP and the GNN ROCs and the GNN shows better performances. You can see in our latest [ update] (slide 12) the compares the two. We will make sure that this plot is also included in the v4 of the AN.
• Section7: L425: this needs to be rephrased. Also perhaps better to move the MC-based estimation and plots to the appendix to avoid any confusions.
  • Agreed! the text will be updated in v4 to reflect the fact that we will be using the DY data driven and reshuffle the plots to make sure it consistent with the text.
• What’s the pT of the third lepton in the 3lep CR?
  • The pT of the third lepton in the 3L CR is added to the pT_miss to emulate the pT_miss in the SR. The distributions of pT_miss in 2lep CR include the third lepton pt. We do not make a selection on the third lepton explicitly but we do on the emulated pt_miss.
• L454: the ptMiss threshold of 120, I guess this is decided based on mZ with some resolutions effects, is it?
  • Both signal and background has falling ptMiss spectra, but the signal falls less fast and the tradeoff between S/B and signal acceptance is favorable for a cut around 120.
• L461: weights are also computed as a function of boson pT, right? Please clarif?
  • This is correct, we do use the boson PT. The re-weighting variables are displayed in Figure 41.
• Figure25: Do we need to keep WW in the 3l CR here? Based on statistics, I’d consider dropping it as it’s adding jumps in the uncertainty band or am I missing something?
  • Yes, we can drop WW in the 3l CR.
• In the SR, why do you have large uncertainty b/w 60-170ish, is it from DY? Actually is the DY in these plots still from MC?
  • Right, the large uncertainty b/w 60-170ish is due to MC DY. The plots with data-driven DY will be shown in later section.
• Figure26: how do you decide the m_jj binning?
  • The binning here is arbitrary, and has no effect on the analysis and the signal extraction.
• Figure 27-28, can you zoom in on the ratio plot to better judge the data/MC?
  • Sure, we will update the plots with smaller range of ratio.
• Figure 35: third plot, there is a clear disagreement b/w 100-200, why?
  • It is probably due to mis-modeling of ptmiss from jets in Wjets MC events. However, we have not looked into detail given that we are using data in this single e-CR, and purity of Wjets in this region is reasonably high (~80%, with 20% contamination from ttbar events).
• Figure36: so 2017 and 2018 detectors are identical and I was expecting misid. Rate would go down there w.r.t 2017. But this is not the case for the barrel, do you understand why
  • The higher midID rate in 2018 is due to the pixel inefficiency around phi of 0.7-0.8 in the barrel region during 2018 data taking.
• Fig41: why is there a peak around 1.2-1.5 for the eta, transition region, effects?
  • Yes. The reconstruction efficiency around the barrel-endcap transition region is low, while in the dilepton events, especially in mumu events, the efficiency is still reasonably OK.
• Fig42: I don’t understand the binning in these plots. I thought you have bin-by-bin scale factors for below and above 120. So is it just one bin below and above 120?
  • we have 2D binning as show in the plot, in pT_ll and pT_miss. The pt_miss has two bins. We have tried various binning, the choice is driven by the need to minimise the error in each bins.
• Fig45: i’s suggest to keep plots with data-drive DY here (third plot, I believe). Do you understand the source of large uncertainties? Figure 47, looks good but I’d be nice to see how input variables look as I pointed out before.
  • Agreed, as stated above we will include these variables in the v4.
• Figure48: why are the impacts from all JES sources one-sided. Low stats leading to huge bin migrations? Also I’m not sure if [NormWW] makes sense. Do we need this CR at all? Actually, could you add a yields table in the SR to understand it better? Pstfit plots in the control regions should be safe to look at.
  • We are still investigating the issues regarding the one-sided uncertainties.
  • We will include the yields in v4 as well as the post-fit plots
• Section 10: Do you have these results using reco level info?
  • Not yet. We are currently focussing on the getting the SM measurement. but will do so for the paper.

Muon object review: Kyeongpil Lee

• According to the answer to the questionnaire, you said that |Muon dxy| < 0.02 and |muon dz| < 0.1 was used, which is tighter than the one defined in Tight ID. But I couldn’t find such statement in the AN, at least in section 3.0.4. Could you clarify which values did you use in the analysis?
  • we inherited the selection criteria from the EXP-19-003 MonoZ analysis. This choice was made because it effectively rejects a larger fraction of fakes while maintaining a high signal efficiency compared to the recommended 0.2 value. In short, this tighter selection is designed to reject non-prompt leptons. We will ensure that the AN is updated accordingly to reflect this statement.
• According to the answer, the correction for the L1 muon pre-firing was not used. Do you have a specific reason not to use it? At least, it should be easy to use as you are using nanoAOD-v9.
  • it seems that there was a mistake when flying the form. We apologise for any confusion—we do indeed apply the L1 muon pre-firing weights directly available from the NanoAOD-v9 trees.
• Just one more following question: If you used tighter dxy and dz selection, did you apply additional efficiency scale factor for that? or, did you do some cross check that the effect of such SF is negligible for your analysis?

Electron object review: Ying An

• Do you apply the Reco SFs for electrons? If you applied, do you separate these SFs into two parts? Since the Reco SFs are calculated separately for pT more or less than 20 GeV.
• Are trigger SFs approved by EGamma POG? Or do you cite them from approved results? If not, could you prepare several slides and give a brief talk in one EGamma meeting?
• For the uncertainties related to the electron, could you introduce how you handle the correlation between years/process in the AN note briefly?

-- HaddadYacine - 2024-02-16