HIN-16-013 Pre-Approval Comments


Comments from Camelia

Q1: General: please make sure (\clearpage?), that every set of figures come around the corresponding text (avoid having a description on one page, and the figure 3 page later).

A1:

\clearpage has been used to separate each section. However, sometimes it is hard to keep figure and corresponding text in the same page.
Q2: Please clarify the triggers you've used? (smth about HFOR efficiency appears in L99, should be moved earlier). Please make more clear what is really new (in terms of selections and efficiency calculations compared) to the previous analyses.

A2:

Changes have been made in the analysis note.
Q3: L61,69: is-->are

A3:

Changes have been made in the analysis note.
Q4: Sec. 2.2: are these privately or officially produced samples?

A4:

These are officially produced samples listed in the official MC production twiki page: https://twiki.cern.ch/twiki/bin/view/CMS/HeavyIonMCProd
Q5:L82 and L83: so we divide pPb DS to pp total-innelastic x-sec. Why is this not a problem, that in pp we are considering also the SD events, while in pp not? Do we still expect the discussion on the RpA to be correct?

A5:

It is the same as HIN-15-015 RpA measurement. pp spectra is corrected to inelastic collision, while pPb is corrected to double-sided events.
Q6:L98,L101-102: why a normalization with lumi means events with <2 tracks do not affect the result?if you are dominated by processes with small ntrack, then your'll just get more of them no?

A6:

Changes have been made in the analysis note.
Q7:Your selection is on Ntrack>=2 (L92). So I don't understand this emphasis. Also, what is the binning in Fig. 21? Meaning, what is the efficiency for Ntrk>=2 (and not bellow).Are your efficiencies the same (have you x-checked) with HIN-15-015? Have you calculated smth differently? Are the numbers approved by the GO group?

A7:

Changes have been made in the analysis note. HIN-15-015 derived event selection as a function of multiplicity. We used number of tracks. These two

should result in similar effect in event selection efficiency corrected spectra. The effect is less than 1% for Ks spectra, as you can see in AN.

Q8:Sec. 3.2: is this an independent study? Are these selections (and efficiencies) different that what used in HIN-12-017?

A8:

No, it is the same as HIN-12-017.
Q9:fig 4: can you make a similar plot with and without the event-by-event corrections for event-selection?

A9:

The required plot has been added to the AN.
Q10:L132 if produced centrally, all MC z-vtx distributions should be the same ...

A10:

Both samples are offical MC. pPb is produced in CMSSW53X while pp is produced in CMSSW75X. Details can be checked in the official MC production twiki page.https://twiki.cern.ch/twiki/bin/view/CMS/HeavyIonMCProd
Q11:Are any of the candiates cuts different from previous analyses? So you tuned anything specifically for this analysis?Why don't you have a symmetric side-band around the peaks?

A11:

The candidate cuts is the same as QCD-10-007. Comparing to HIN-15-006, the only cut difference is whether to use cosine pointing angle or DCA significance cut for Ks and Lambda. The asymmetric side-band is purely because we want to place the peak and text in the same plot. Furthermore, as checked in HIN-15-006, changing fitting range will result in spectra change within 1%.
Q12:L154: are you assuming some sort of convention, by "-" you actually mean both charge conjugates (case in which please clarify), or you really look only at negative charge only?L155: add space before "(". L156: might be worth writting down the decay chain that you are reconstructing.

A12:

We actually mean both charge conjugates. It is mentioned in introduction part L51. Decay chain has been added to the AN.
Q13:L176-182: So all parameters are left free in the fit, or you fix smth to the MC? Is this a binned or an unbinned fit?

A13:

For invariant mass peak fit, RooFit package. Fitting is performed with MINOS. All parameters are left free and it is a binned fit.
Q14:Fig. 13-20 (also 25-32): please add a pull distribution. It will make it easier for you too to see failed fits (like there seem to be in the first bin showed in most of the figures, in which the right tail is missed).

A14:

As mentioned in A13, RooFit package is used for the fitting. Fitting is performed with MINOS, which can give uncertainty of fitting parameters with correct covariance matrix. In both QCD-10-007 and HI-15-006, we don't perform the pull distribution. However, chi2/dof is added in the invariant mass peak fitting plot, which can show the goodness of fitting. Also, the potential uncertainty in the fitting can be tested by changing the background function form, which is one of the systematic uncertainty sources that we have checked.
Q15:L265: remove the "/" and no need to redefine Delta_R. L259 and 266: can you show a distribution of this distances, in both data and MC?

A15:

Delta_R is a variable used to perform matching, meaning you can not get Delta_R in data, since it requires gen part. Delta_R distribution of MC has been added to the AN.
Q16:L325: is this different than what was used for HIN-15-015 (the RAA/R_pA paper)?

A16:

Previously, we used the number 3.9% from CMS-PAS-TRK-10-002. Now, we use the same value as HIN-15-015.
Q17:It is not clear, when there is no pt-dependence, how you chose the final value to propagate; alternatively, when you make a pT-dependent uncertainty, how you chose the 'bin-by-bin' final uncertainty. You say vs pT, yet you talk also about low-pt vs high-pt, as if only 2 values were used for the 'pt dependent uncertainties. I suggest some sort of stacked-histos summary for all these uncertainties, vs pT, to show individual contributions, if instead it's more than listed in tables.

A17:

If there is no pt-dependence, the final value is choen so that almost all the points are within two lines: 1+syst. and 1-syst. For this analysis, we only observed some sources may have different values for pt< 1GeV and pt> 1GeV, due to the fact that background level in pt<1GeV is higher. So for those sources, we treat pt<1GeV and pt>1GeV separately, with the method mentioned above. That is why we only have 2 values for pt dependent uncertainties. Following your suggestion, stacked-histos summary plots have been added to the AN.
Q18:Please fill the stat. comm questionaire --> some of these variations are checks, and hence should not be used as sytm. uncert.

A18:

Stat. comm. questionaire has been filled.

Follow-up Comments from Camelia

Q0:can you please add to Twiki where you have 'changes have been made in the AN' exactly where the changes were done, and what exactly figures were added? (going again through 126 pages is not really optimal)

A0:

Sorry about that. More details are shown in the following answers. The current AN is version 6.
Q1:Fix the figures reference in text ("??" in many places)

A1:

This happens to fig1 to fig5. Now it has been fixed.
Q2:The 'it was done before' is not enough. Can you give an estimate of what fraction of total innelastic collisions (which you use in pp) is coming from DS events only? Is there any pT dependence in the other contributions (single diffractive, non-diffractive, central-diffractive)?

A2:

We don't have an estimate in pp. But for pPb, the GO group did the study here: https://twiki.cern.ch/twiki/pub/CMS/HiCentrality2017/170425_DoubleSided.pdf . From slide 2, the fraction of DS events over total inelastic events is about 2-3% from HIJING and AMPT (EPOS numbers also available in the table). So if we correct pPb DS to inelastic, the RpPb will change 2-3% (become smaller). There are pT dependence for contributions from different processes. There was a study here for charged particles on slide 3: https://twiki.cern.ch/twiki/pub/CMS/RpPb2015/spectraPP_ST_Mar16v1.pdf . However, the effect should be negligible since the missed events by DS selection are only a few percent, and more importantly there is almost no track (strange particle) in these events. We are using DS selection to agree with previous publications (spectra in HIN-12-016, HIN-15-006 and RpPb in HIN-12-017 and HIN-15-015). In HIN-12-017 (https://arxiv.org/pdf/1502.05387.pdf), on page 10, there was a discussion says: "As double-sided events correspond to 9497% of inelastic collisions based on HIJING and EPOS MC computations [31], the value of ⟨TpPb⟩ would be about 5% higher for double-sided events." We prefer to keep using DS in pPb to avoid introducing inconsistencies within CMS. We will make it clear in PAS that we are using DS in pPb and inelastic in pp.

Q3:A7: so you derived or not your own efficiency (vs Ntrack)? Not clear if they were OK-ed by the GO group also... a quick check would be for you to re-make these efficiencies vs the same observable as HIN-15-015 and see if you get the same results in their variable no? Also, I don't grab the difference between eff vs reco tracks (15-015), vs eff vs number of tracks (your case). I mean they also have N_tracks (which are number of reconstructed tracks)... So please clarify where is the difference.

A3:

To avoid any confusion. Now we are using exactly the same variable "multiplicity", as what HIN-15-015 used, which is number of tracks that pass track selection criteria listed in L103 to L105. And the correction is found to be less than 1%, which is consistent with what HIN-15-015 observed.
Q4:It's still not clear in Sec. 3, what is used in previous analyses and what not. Please add a sentence in each sub-section, clarifying this.

A4:

The corresponding sentences have been added. For V0, they are in L186 to L189. For Xi and Omega, they are in L220 and L221.
Q5:Fig. 3: you have #tracks on the left, and Multipicity on the right on the x-axis. What's the difference? (and both different than x-axis title in Fig. 5; please make consistent).

A5:

Now they are consistent. Multiplicity here means number of tracks that pass track selection criteria. Track selection are listed in L103 to L105 for pp and L128 to L130 for pPb.
Q6:Q9: is this figure 5 in the AN? If so, is the caption correct, this ratio includes both trigger and ev sel efficiency?

A6:

It is Fig4. It compares Ks spectra with both trigger and event selection efficiency correction to Ks spectra without such correction. The caption is correct.
Q7:Q11: The 'The asymmetric side-band is purely because we want to place the peak and text in the same plot' is not a strong reason. Please switch to a symmetric side-band. Assuming example of fit peaks will be added in the paper too, this question will come up right away from anybody, and can not return the same answer.

A7:

The analyzers agree with this comment. Now the side-band is symmetric, which can be found in Fig64.
Q8:Q14: RooFit is a tool, not also a measuring tape for the goodness of the fit. I can also point you to many analyses in which the pull distribution is added. The chi2/ndof (as also mentioned in by the StatComm reviewer) is not bulletproof and can nto tell you by itself if you don't have smth pathological wrong in the tails. Add the pull distributions in all figures (and by pull I mean https://cms-results.web.cern.ch/cms-results/public-results/preliminary-results/HIN-16-023/CMS-PAS-HIN-16-023_Figure_001-a.png ; RooFit manual has example on how to produce such things).

A8:

The analyzers agree with this comment. Pull plots have been added to the invariant mass peak fit. One example can be found in Fig15.
Q9:A15: which figure?

A9:

Fig33 to Fig36 show the delta R and delta L distribution. This can only be done for MC, since Sim is required for matching.
Q10:A17: which figures?

A10:

The plots are placed in systematic section. One example of such plots is Fig 73.
Q11:You were requested at the preapproval to separate checks from systm. uncertainties. Can you point out to the conclusion of this?

A11:

From the analyzers, we think "rapidity binning" and "beam direction" can be treated as checks. However, we would like to discuss with ARC before making any change to current systematic tables.

Comments from Wei

Q1: Please make it clear (with equations) in AN whether spectra are normalized by cross section or # of events for pp and pPb, respectively. As pp data are normalized by cross section, please provide the value of normalization. In the figures, L_int should be values corresponding to the actual MB sample, instead of full delivered data sample.

A1:

The corresponding changes have been made in AN. We agree that L_int should be values corresponding to the actual MB sample. For pp, we can get L_int easily with brilcalc. However, for pPb, lumicalc2.py needs to be used. We will ask expert how to obtain the L_int for pPb MB later.
Q2:For R_pA result, T_pA instead of Ncoll and pp cross section uncertainties should be quoted. For spectra results, the legend should clarify that pp spectra is normalized by 70mb. In general, please follow consistent presentation as previous R_AA and R_pA paper such as HIN-15-015.

A2:

The corresponding changes have been made in AN.
Q3:Bumpy structure is observed in efficiency vs pT. If due to statistical limitation, please smooth it out to avoid such structures in the corrected spectra.

A3:

Now we smooth efficiency with pol4 as what is used in HIN-15-006. The bumpy structrure looks better now.
Q4:Tracking efficiency uncertainty for Y_asym should be canceled, while should not be for R_pA. In general, in AN, make it clear what uncertainties are considered canceled.

A4:

Systematic tables have been updated. Tracking efficiency uncertainty for R_pA is partially cancelled. For now, we quote 8% for V0 R_pA and 12% for Xi and Omega R_pA. The exact number to quote for R_pA tracking efficiency uncertainty will be discussed in ARC meeting.
Q5:Clarify in AN how the value of systematic uncertainty for each source is determined from the ratio plots.

A5:

If there is no pt-dependence, the final value is chosen so that almost all the points are within two lines: 1+syst. and 1-syst. For this analysis, we only observed some sources may have different values for pt< 1GeV and pt> 1GeV, due to the fact that background level in pt<1GeV is higher. So for those sources, we treat pt<1GeV and pt>1GeV separately, with the method mentioned above.That is why we only have 2 values for pt dependent uncertainties. Stacked-histos summary plots are also added to the AN.
Q6:Show a plot of nonprompt Lambda fraction in EPOS with cuts applied in this analysis.

A6:

The plot has been added to AN.
Q7:Please fill statistics questionaire.

A7:

The statistics questionaire has been filled.
-- HongNi - 2017-05-24
Edit | Attach | Watch | Print version | History: r6 < r5 < r4 < r3 < r2 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r6 - 2017-06-02 - ShengquanTuo
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2021 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback