Analysis Note:
http://cms.cern.ch/iCMS/jsp/db_notes/noteInfo.jsp?cmsnoteid=CMS%20AN-2014/105
CADI/PAS:
http://cms.cern.ch/iCMS/analysisadmin/cadilines?line=SMP-14-020&tp=an&id=1413&ancode=SMP-14-020
ARC Comments through approval of PAS:
Wbb8TeVARCResponses
Version sent to EPJC
l 222 “efficiency correction factors” > “corrected efficiency” otherwise is not a cross-section
What would be the prediction of the
presently used models/calculations, when ran for the 7 TeV? I guess something
in the line of 0.45 pb, while the measurement was 0.53 pb. Still only one sigma or so too
high but the same tendency as the data in this analysis.
)
L181: why do you take a 50% uncertainty on the QCD? where does that estimate come from? Is it just some random "large" number?
Table 2
a) The numbers in the 'muon, initial' column do not add up to the data number.
L130-131 seem inconsistent with L86-87. In L86-87, the shape of the Wbb is taken from dedicated W+bb sample, while L130-131 says the shape is taken from W+jets samples.
Paragraph at L130 describing jet flavor definition in simulation:
"Jets with distance parameter smaller than R=0.5 with respect to lepton are removed from the event." This is done in order not to count prompt leptons as genjets. So leptons from b-hadrons should not be considered in the procedure otherwise it will remove jets with semileptonic b decays. This indeed is stated in the sentence "Generated leptons originating in simulation from the decay of b-hadrons or tau-leptons are not considered." However, it's unclear why leptons from tau decays should not be considered in the procedure. Does it mean it's ok if they get reconstructed as genjets? Why don't simply state that "Jets with distance parameter smaller than R=0.5 with respect to prompt leptons from V decays are removed from the event."
L 150-153: Is the QCD normalization extracted in the MT<20 GeV region the same as the one you get when you use the anti-isolated region?
- l.158: "on the third" -> "on a third"
lines 228-229 11% cannot result from 80% times 40% … what you
probably have to clarify is that there are these two contributions,
but also the geometrical acceptance.
| norm. + shape | correlated
| Norm. + shape | Correlated
), the analogous W+b final state can not and makes a negligible contribution to the signal region. In contrast, W+cc where the cc is produced via gluon splitting has very similar kinematics to the signal region: W+bb. So the lack of distinction between W+b and W+bb at the generator level was chosen so as to be as inclusive as possible while the distinction being made between W+c and W+cc comes from the fact that both final states occur at a nonnegligable rate but arise from different processes.
I found the entire paragraph L 161 – 168 a bit confusing. I also am not sure if the shapes referred to in L 161 is the eff_b or JES shapes or both?
These "shapes" referred to the shapes of the mT variable in the simulated samples (MC and QCD).
Had:
First, the shapes obtained from the simulation together with the QCD shapes are fitted to $\MT$ data distributions in the $\ttbar$-multijet region. The obtained b-tagging efficiency rescaling factors are measured separately in the muon and electron control samples and averaged before being applied. The reweighted simulated samples are used in the next step, where the $\ttbar$-multilepton region is used and the jet energy scale in the simulation is adjusted. As a result of these two steps, the simulation is expected to properly describe the $\ttbar$ contribution and can be used to extract the number of $\Wbb$ events from a fit in the signal region.
Became:
First, using the simulated samples detailed above,
a fit was performed in the $\ttbar$-multijet reigion
using the $\MT$ variable.
The result of this fit gives an estimation of the
b-tagging effeciency rescaling factor, which
is measured separately in the muon and electron
channels and averaged before being applied.
The reweighted samples are then used in the next step
where a fit to the $\MT$ variable in the
$\ttbar$-multilepton region is performed and
the jet energy scale in simulation is adjusted.
As a result of these two steps, the simulation is
expected to properly describe the $\ttbar$ contribution
and the final step is to extract the number of
$\Wbb$ events from a fit in the signal region.
L178 – 184: I don’t understand how putting the uncertainties at 100% allow the fit to remeasure them. Or maybe I do. But I don’t understand what is going on there. Are these backgrounds essentially allowed to float within a 100% uncertainty?
Yes, this is correct. Because we are remeasuring these scale factors whose effects are very difficult to disentangle, our strategy was to adjust the simulation according to the fit, but allow future fits to recover the initial scaling at 1 sigma. If we saw that successive fits pulled the scale factors substantially back toward their original value, this would be evidence of bias in our method, but we see instead that the rescalings stay fixed. This is also confirmed by performing closure tests, re-performing a fit after extracting and applying the scale factors and seeing that it then delivers a scale factor consistent with unity.
And I am curious about the asymmetry between allowing the b-tagging and JES uncertainties to
float between 0 – 200%, while the PDF uncertainties are a factor of two (50 – 200%)? This asymmetry confuses me.
For the b-tagging and JES uncertainties, the uncertainty was chosen specifically with the intention of setting particular values as being one standard deviation away from the central value (namely, the value that "undoes" the rescaling). The uncertainty associated with the change in scale is estimated by changing the renormalization and factorization scales up and down by a factor of two, but the final value is ~1%. The PDF uncertainty is estimated by taking the bounding window of estimations from a variety of PDF sets, and is ~10%, thus giving the combined PDF+scale = 10% given in Table 1
I understand the equation on page 7, give or take. There is an implied “ = “(N^Data_signal / N^MC_signal) x (N^MC_generated / Lumi)” step, or at least so I think. The first term is taken to be the signal strength, while the second term is sigma_gen. I think? I have never heard the term “signal strength” and am confused by it. In addition, It is clear what you are substituting in when you equate the second and third terms. But if I understand that properly...and I probably don’t...N^MC_signal includes acceptance and efficiency, but it also has a background contamination. I guess N^Data_signal also has a background contamination, and it looks like you are assuming they are the same. I think?
You are correct both in the missing equality and in the problem that one encounters when trying to use this equation directly substituting N_events for the different variables, and this is what motivated us to use signal strength in the first place. In performing a fit, the final number of signal events will change as a result of two sources: the effect of uncertainties which are correlated across different samples (things like b-tagging, JES, MES, EES, ...) and the effect of a change to the cross section of the signal sample itself. Factoring the total change in signal normalization into these two components then, the signal strength is just the component coming from the change in cross section, and thus is applied to the generated cross section used to find the measured cross section as indicated in the equation.
L228 & L227. I don’t know how to reconcile the 11 (13)% and the 80% and 40%. (Because isn’t 80% x 40% = 32%?) The missing piece here is that 40% is the btag efficiency per jet, so the calculation is 80% x 40% x 40% = 12.8% which is consistent with the 11/13% quoted above.
To get 11/13% we invert the equation on P7 and the numbers in Table 2 acc x eff = ( n_wbb ) / (lumi * xc * ss) to give mu: 0.1329 ele: 0.1148 We estimate the efficiency of the lepton requirements and (per jet) b-tagging by looking at the raw number of events passing our selections
Added statement "per jet" to Paper
L252 – 253: you say that it agrees with the 7 TeV Z+b analysis. I know you give the reference, but it would help the lazy reader if you were to quantify your statement here.
added
I do not think that we have user defined symbols in the abstract. For example, pTlep is given as $p_{\mathrm{T}}^{\ell}$ . If there exists a list of allowed symbols, that would be useful, but I have not found such a list.
done
L6: I’m not sure that you need to spell out the acronyms of the two experiments. While it’s not wrong, it’s ugly and jarring.
done
L9: Remove “(minimal or not)”
done
L10: through -> using
done
L11: hadrons -> hadron^M
done
L12: add comma: boson,
done
L 13 – 14: replace “to bjets as jets which originate from hadronization of b quarks” with “to jets which originate from the hadronization of b quarks as b jets”
done
L 17: “In the past, the” -> “The”
done
L18: Remove “by the ATLAS and CMS collaborations,”
done
L19: Remove “previously”
done
L 28 & L35. Shouldn’t pseudorapidity be defined?
defined
Equation 1: Why I^{rel} and not just I?
changed to I
L65: that -> which
done
L72: “Tight” is jargon. Drop and rewrite as “both jest pass a threshold which has”
done
L77: You say “signal region,” which seems a little jargon-y to me. Maybe “signal sample”? I’m a little iffy on this suggestion.
changed
Had:
After all signal region selection requirements are applied,
Became:
After all selection requirements for the signal enhanced dataset are applied,
L95: You define diboson here and yet use it on L79. Maybe define it earlier?
defined on L79 now
L100 – 102: “The ttbar...measurements [42].” Is a bit unclear. Maybe “The ttbar cross section was determined from data [42] using ATLAS and CMS data.” ????
how about:
The ttbar cross section was determined from data colleced by the ATLAS and CMS detectors [42].
L113: “fit, except QCD, are taken from simulations” -> “fit are taken from simulations, except QCD.”
done
L145: You say “all other backgrounds.” That means “all backgrounds not yet mentioned,” right?
No, this means all backgrounds other than QCD. We apply the QCD selections to Data and MC, then take the difference as the QCD shape.
L151: affect significantly -> significantly affect
changed L159: you explicitly call out LO and I see how that is relevant in context. But I don’t see where in the background calculation discussion where it was called out earlier. It seems to me that that should be done. Both the 4F and 5F samples are generated with MadGraph as mentioned on L82. No change made - ok?
L161 & L166: It says three steps and two steps and I imagine that this means that the steps are broken up into 1 + (2 + 3), but I can’t see how that was done.
It isn't. This was just a description of steps 1 and 2. Step 3 is the fit in the signal region described in the Results section.
L 161 – 166: I am having trouble understanding the sentence “First, the shapes...multijet region.” Please clarify. Does this fit include the W + bbbar?
Agreed, this was awkward. The intention was to say that everything was fit (but QCD comes from a data driven technique, so I didn't want to call it simulation)
The whole paragraph changed to:
The fit procedure thus consists of three steps.
First, using the simulated samples detailed above,
a fit was performed in the $\ttbar$-multijet reigion
using the $\MT$ variable.
The result of this fit gives an estimation of the
b-tagging effeciency rescaling factor, which
is measured separately in the muon and electron
channels and averaged before being applied.
The reweighted samples are then used in the next step
where a fit to the $\MT$ variable in the
$\ttbar$-multilepton region is performed and
the jet energy scale in simulation is adjusted.
As a result of these two steps, the simulation is
expected to properly describe the $\ttbar$ contribution
and the final step is to extract the number of
$\Wbb$ events from a fit in the signal region.
L173: add comma: cases,
done
L174: I think the “ prior to “norm” should be ``.
changed
L175: add comma: procedure,
done
L191 – 192: I don’t understand in detail the implications of this sentence.
It means that we rescaled the simulation using the scale factors described in the previous sentence. Removed "accordingly".
Table 1, [caption]: what do you think about changing UNC for X? I spent a lot of time wondering about what uncertainty UNC was.
changed
Also, sentence starting in line 4 “In the ‘variation’...UNC in the fitting procedure.” Seems to have an extra word or something that makes it not entirely clear to me.
Not sure what to do - didn't change anything
Table 1: JES rescale line & L196: I’m confused by the 1.3 & 1.6 disparity.
1.3 is correct, changed.
I’m also a little confused by how these tie into the “in the fit” on L194.
Clarified by removing the statement: The energy scale in the fit varies within its uncertainties, $\sigma_{\mathrm{JES}}$.
Figure 2 [caption] vs. figure 2 [figure]: You use different terms, specifically “ttbar- multilepton phase space” vs. “ttbar multilepton control region”. Personally, I think it might be best to use the phrase “ttbar multilepton enhanced data set.”
changed to ttbar multilepton enhanced data set in both places
L213: There is an extraneous “)”
removed
L240: add comma: scheme,
added
You spell out LHC and CMS but
not ATLAS changed
l13 note -> paper changed
l22-23 for the paper you need to add the detector section back to the text. added as directed in https://twiki.cern.ch/twiki/bin/viewauth/CMS/Internal/PubDetector
All figures: should now have CMS without Preliminary. changed
Things we can probably discuss in parallel to the LE review:
New fig 4: could rebin the lepton pT distribution above pT>100 GeV ? Or
show in log scale ? agreed
For the choice of distributions to show, I don't know if we want to
limit ourselves or show a representative set for the kinematics of both
jets and lepton, or search variables like Mjj ? Should we have an
ARC-author meeting about this in parallel with the LE review ?
possible distributions are shown above
Comparisons between the results of calculations performed
under different assumptions provide important feedback
on the functioning and validity of the techniques employed.
Differences in predictions arising from the modelling of
b quarks as massive or massless are possible, as are
variations in predictions arising from the use of different
showering packages (\PYTHIA6 vs. \PYTHIA8) or matrix element
generators (\MADGRAPH vs. \MCFM).
In the phase space explored here, these predictions are all
very close in their central value and agree with each other
well within their respective uncertainties. 2) You describe around line 81 that "the 4F signal sample is used for the shape.. the normalization from the 5F W+jets sample". It is described quite clearly how the 5F W+bb signal is defined from the W+jets samples, but not so much how it is acutally merged with the 4F sample. E.g.: in my last reading I was wondering if you do you apply the same gen-level selection criteria on the 4F W+bb sample to define this sample. As this merging of two samples is not trivial to do consistently, it might be helpful to give a few more details. The 4 flavor sample is generated as W+bb, and then the same reco-level signal region selections as are imposed for any other sample. To "merge" with the 5F sample, we simply rescale the yield of the 4F (after all selections) to match the yield of the 5F (after all selections). 3) You have now added Fig.4, with some kinematic distributions, which is nice. However, the description is currently still quite minimal (and as AM said: we could discuss which variables to show). The technique for obtaining QCD is slightly different for these variables than for mT, but the procedure is otherwise the same (hence not so much of a description for how we obtained the distributions). For QCD, the difference is that instead of using an inverted isolation sideband for the shape, we use an mT<30 GeV sideband. Then the normalization is set to be that used for the mT variable. As a suggestion, it would e.g. be nice to discuss the data/simulation difference around DeltaR (b,b)~3.5. Note that such a tension was also observed in previous CMS measurements of Z+bb. (E.g. SMP-13-004 and EWK-11-015, but unfortunately I cannot upload attachments in HN now.) Looking at https://cms-physics.web.cern.ch/cms-physics/public/EWK-11-015-pas.pdf Figure 5 does not show a difference at dR(b,b)~3.5 while some tension is seen in AN2012_303_v7 (I couldn't find dR(bb) distributions in any of the papers) and there the comment is made that this "suggests a harder spectrum in data than expected." In general, the agreement between simulation and data in this analysis is good, and it is hard to draw a conclusion from one data point being 20% high, especially when the points to either side agree at the 10% level. This could be statistics or it could be QCD, but it is not clear what conclusion you would like to be drawn here.
CADI/PAS:
http://cms.cern.ch/iCMS/analysisadmin/cadilines?line=SMP-14-020&tp=an&id=1413&ancode=SMP-14-020
ARC Comments through approval of PAS:
Wbb8TeVARCResponses
Comments from Referee EPJC
Full commehts
Version sent to EPJC
1. Sec. 4: "For the signal distributions, the W+bb component of an inclusive
W+jets sample is used, with the shapes of the distributions taken from a
dedicated high-statistics generated sample of exclusive W+bb." It is not completely clear to me what you do
here: do you take the W+bb subsample of the inclusive W+jets sample, and then reweight distributions of
these events based on those extracted from the exclusive W+bb sample? If so, I would write that more
explicitly then.
We had two samples: one which was W+Jets (5F) and one which was W+bb (4F). The first of these was used for the cross section and normalization and the second was used for the shapes.
I think he is right, we need to change what is written and anyway we should not put 5F and 4F in the answer to not mislead him. So I suggest:
For the signal distributions, the W+bb component of an inclusive W+jets sample is used, by separating the W+jets simulated sample into three subsamples labeled as W+bb , W+cc , and W+udscg, which are defined below. .... explaining the method ...
Leptons that do not originate from the primary vertex are not considered for selection. To further improve the shape description of the Wbb distributions a dedicated high-statistics generated sample of exclusive W+bb is used, while keeping the normalization of the distribution as obtain from the procedure explained above.
2. Sec. 5 par. 6: "allowed to vary within the uncertainties" - If you add Gaussian constraint terms to the
likelihood for these uncertainties, I would explicitly write so, here or in sec. 6.
Yes not a strict cutoff
Do not understand your answer, we can add something like - using log norm distributions ...
3. Sec. 5 par. 8: Did you consider an option of simultaneous fitting of the signal and two control regions?
Naively I would suppose this might improve the overall precision a bit; if this is not the case (or you
don't want to use it for some reason), I think it's worth commenting in the text.
We did try this first, but the shapes are not controlled, etc etc...
Yes we also performed a simultaneous fit and found that it gives worse precision since there are component in the fit that work in different directions, thus compensating each other, that makes fit less stable. Instead we perform step-by-step fit where each component gets some predefined values that at the
end help fit to stably converge. Suggest to end at the end of this paragraph. "Similar results can be obtained by performing a simultaneous fitting of the signal and the two control regions. It is found that the b tagging efficiency correction and JES correction have similar effects on the distributions and thus compensate each other effects in a simultaneous fit that reduces its precision. Separating these effects in steps provides better understanding of underlying uncertainties and therefore
more precise results."
4. Fig. 2: I am not sure you have a clear message with this figure. If by this you pretend to have studied
the kinematical properties of the Wbb system, at least some statement on the agreement with simulation is
needed.
We see good agreement in angular separation between well separated b jets and lepton transverse momentum within 10% until 100 GeV with a general falling trend
You need to suggest some text here as I do above
5. Sec. 7 and Table 3: It is not obvious how the uncertainties listed in Table 1 are broken down into
(syst) and (theo), e.g. where go the uncertainties of the sample cross-sections? I suggest to clarify what
"theo" includes when quoting the final cross-section number.
Sample cross sections are in syst
please write more extended answer
Comments for Final Reading (on Paper V9)
Sijin Qian, August 8
In general
(1) The Abstract in Page 0 and the Conclusions Section in Page 11. It
is quite surprising that the Abstract and Conclusions Section in v9 are
identical; this seems very unusual (if not never) in all 500+ CMS papers
published so far. Would you please consult with other CMS papers to see
whether you would like to be conformed with them (to avoid this kind of
identical)?
- These were different at one point, but I was advised to change and make them the same. We can reword the conclusions, to be discussed during FR
=
= In v9, thank you for having changed two places; however, L13 has
= a new case of "a b quark", and it will be appreciated if the same
= change can be made.
- all instances of "a b xxxx" have been changed
- changed
- explained (and merged with another comment)
- gap has been reduced
=
= In v9, thank you for having changed the 7th lines; however, the
= 2nd lines (i.e. "Fit Uncertainty") in all plots of Figs.1 and 2
= still need to change the word of "Uncertainty" to the lower case.
- changed
- changed
=
= In v9, thank you for having changed for the suggestion (b); however,
= the suggestion (a) seems have not been touched yet, i.e
= two words of "number" may be saved as
=
= "where Ndatareconstructed is the number of observed signal events,
= NMCreconstructed is the number of expected signal events from
= simulation reconstructed in the fiducial region,
= NMCgenerated is the number of generated events in the ...," -->
=
= "where Ndatareconstructed is the number of observed signal events,
= NMCreconstructed is of expected signal events from
= simulation reconstructed in the fiducial region,
= NMCgenerated is of generated events in the ...,"
- omitting the word "number" sounds awkward, will defer for FR
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
Darien Wood, August 8
Dear SMP-14-020 authors,
If found the paper to be clear and easy to follow. My comments on draft 9 follow.
Best regards,
Darien
Type B:
1. I would have found it interesting to include the separate results for the different W charges: W+bbbar and W-bbbar. This separation has more physics interest than that between the electron and muon decay modes, which are already included.
- Unfortunately the result can not be changed at this stage.
- Yes, this difference is negligible, especially given that we have a veto on third jets.
- simulation of QCD with Pythia6, added to text
- we have gone through many iterations on this statement, if we try to say "almost not effected" or "minimally effected" people ask for the size of the effect and it is 0, therefore
- Correct, we have modified to make it clear "The uncertainty due to the b tagging efficiency and the uncertainty due to the JES are observed to only affect the normalizations of the samples in the \MT variable."
- yes, see above
- we were asked at one point to mention this study, and no change has been made to the text on this point
- these are the four components stat/syst/theo/lumi, caption now reads " The measured cross section is also shown with the total uncertainty in black and the luminosity, statistical, theoretical, and systematic uncertainties indicated."
- changed "which" -> "that", was asked to include "decays leptonically"
- changed to "Throughout this paper, a hadronic shower originating from a bottom or anti-bottom quark is referred to as a b jet, and b-tagged jets are the reconstructed objects either in simulation or data that have been identified as such."
- changed
- we agree, but were also explicitly told not to hyphenate
- line changed
- we were asked specifically to include this at one point
- changed to "resulting distributions"
Kajari Mazumdar, August 8
Dear Authros and ARC members,
Congratulations for the interesting work. Please find my comments below for
the FR version.
I am very sorry about the delay in sending these.
Kind regards -- Kajari mazumdar
Type B
Throughout the paper, the order for muon and electron needs to be
exchanged. Any reference to electron should appear before that of muon, to
maintain the natural order.
- have been told to keep mu first, but consistently in all the paper
- 40% is for our selections
- We have it included in the selections, but it isn't used in making mT.
- text clarified
- was asked specifically to put this in at one point
- removed
- it can be done in FR
- changed
- we assume the reader is familiar with vector bosons - when we introduce the notation "V" we define V = W or Z
- suggest to discuss in FR, fine with us though
- The properties of the jets that are made in association with the boson, changed "dynamics" to "kinematic properties"
- changed to " a hadronic shower originating from a b or anti-b quark is referred to as a b jet, and b-tagged jets are the reconstructed objects either in simulation or data that have been identified as such."
- added
- added "the definition of the isolation variable.."
- changed
- added
- did not add "considerably" to statement "long lifetime"
- changed
- this is where the signal selections are detailed
- added "background"
- changed
- was asked to include it at one point
- kept in present tense, according to PubCom rules
- this formulation is the result of many iterations of comments
- We have delta R defined, and changed "R" to "delta R"
- added
- changed to "If an event does not contain a b jet"
- definitions of the two regions and their naming scheme is given in the previous paragraph
- changed
- changed
- changed
- multijet - changed to "Because $\ttbar$ production may have more than two jets in the final state, the rejection of events with a third jet makes it sensitive to JES."
- removed
- changed
- changed
- changed to "The renormalization and factorization scales respectively are set at $\mu_{\mathrm{R}}=\mu_{\mathrm{F}}=m_\PW$, and the uncertainties on this choice are estimated $\mu_{\mathrm{R}}$ and $\mu_{\mathrm{F}}$, up and down by a factor of two."
- changed
- Z2* tune kept
- changed
- removed
Michael Schmidt, August 7
Type-B
line 130-1: Does your consideration of these two control regions
lead to a consistent correction? If so, perhaps you should state
that fact since it is not automatically true.
- yes, we have checked this with closure tests, but did not modify the text
- yes, we require dR(lepton,jet)>0.5 and this was included on L98
- We are taking max(unc1, unc2) and this line has been added " with the combined uncertainty taken as the maximum of the uncertainties on the individual lepton channels."
- Ref 55 has been updated, ref 46 is the combined cms/atlas result from papers ref 47,48. I couldn't find the combined result as a paper - Sascha?
=========================================================
Type-A
There are several instances in which a number is separated from
its unit by a page break. This makes the text slightly harder to
scan. I find it worthwhile to connect a number to the unit using
either tilde or \mbox. So, in the Abstract,
\sqrt{s} = 8~TeV or \mbox{\sqrt{s} = 8 TeV }.
You can also do this for W~boson in the title, for example. I admit this is a cosmetic consideration. - changed
- changed
- we were also asked to include leptonically, we suggest to discuss in FR
- was also asked to include leptonically, the same as above suggest to discuss in FR
- changed
- removed
- I agree but was asked to put this in at one point
- changed
- kept so as to not start sentence with a symbol
- changed
- changed
- changed
- changed
- added
- removed
- changed to "resulting distributions"
- added
- changed
- changed
- added "samples"
- changed
- removed
- changed
- not changed, can be discussed during FR if necessary
- definition occurs only once now, the first time mentioned
- changed
- not changed, suggest to discuss in FR
Steve Wasserbaech, August 6
> Type B
I'm not sure what to propose, but I'm wondering whether it is appropriate for us to say we are measuring the W+bbbar cross section when we don't distinguish between b and bbar jets...
- the title says "in association with two b jets"
- changed
- changed
- ok, changed
- cross section of a given process - what do you recommend? Can be fixed in FR
- changed to "a"
- changed to "that", and we have been told specifically to include this information by other commenters, lets discuss in FR
- removed
- removed
- changed "requires" to "contains"
- changed
- Yes, gen-level b are called b jets, and reconstructed objects are b-tagged - changed to "Throughout this paper, a hadronic shower originating from a bottom or anti-bottom quark is referred to as a b jet, and b-tagged jets are the reconstructed objects either in simulation or data that have been identified as such."
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- removed "events induced by"
- changed (but had been told at one point that we should capitalize)
- changed
- changed
- removed
- changed
- ok, removed
- added
- changed
- changed
- changed
- changed to "contains"
- changed
- changed
- changed
- added
- changed
- changed
- changed
- changed
- changed to distributions
- added
- changed
- changed
- changed
- changed
- changed
- added
- changed
- changed
- added
- changed
- was asked to change to "10\% to 15\%"
- changed
- removed, just validity
- changed
- changed
- changed
- changed
- changed
- not sure how to include in bibtex file, presently as : collaboration = "ATLAS, CMS",
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
Dick Loveless, July 26
Abstract:
line 1: Change "which" to "that" and delete the comma after "leptonically" since it is a restrictive clause. Maybe leave out this phrase since it is repeated on line 5.
- which->that and comma deleted, phrase kept
- removed
- changed
- removed
- removed
- removed
- changed
- changed
- changed
- it is looser than our offline selections, do we need to be more specific
- changed
- changed
- changed
- words changed - was asked at one point to take out mT definition
- changed
- changed
- changed
- %GEERN% removed
- changed
- changed
- Nothing changed yet. Could just start with "Backgrounds to the W+bb..." and remove reference to later section
- changed
- changed
- changed
- removed
- changed
- changed
- changed
- changed
- changed (with comma after "In the simulation")
- added
- added
- removed
- changed
- changed
- changed
- changed
- To avoid pronoun "we", changed to "Therefore, the data and simulation are compared .."
- changed
- changed
- Not changed. The idea was to put signal and ttbar selections all in the same place and not repeat. Maybe rename Section 3 as "Event reconstruction" and Section 5 as "Event selection and analysis strategy"
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- added
- changed
- changed
- added
- changed
- added
- changed
- changed
- changed
- added
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed - Scale factors are provided by the POGs and applied, but are remeasured for this analysis, hence "rescaling" %ENDCOLOR
- added
- changed
- added
- changed
- changed
- yes, changed
- added
- changed
- changed
- The PubComm guidelines seem to recommend against this : "do not put a horizontal line at the very top or bottom of a table" - last paragraph in https://twiki.cern.ch/twiki/bin/viewauth/CMS/Internal/PubGuidelines#Figures_and_tables
- changed
- changed
- changed (to "Based on the fits...")
- changed
- The initial distribution is obtained after applying the results of the ttbar fits on the Wbb simulation. The box seems to be recommended against.
- changed
- changed
- good point, changed
- changed
- Again, not changed
- changed
- changed
- changed
- changed
- changed
- removed
- changed
- changed
- removed
- changed (except 341 which is pp)
- added
- changed
- changed
- added
- changed
- Not changed, but ok with us, lets discuss in FR
- We measured the cross section with the leptonic decay of the W, but did not measure the branching fraction ourselves. If interested, theorists can easily do this multiplication
- changed in all the places indicated
- changed to "The measurement of $\PW$ or $\PZ/\cPgg^*$ (henceforth denoted $\PZ$) production in association with b quarks in proton-proton collisions provides important input for refinement of calculations in perturbative quantum chromodynamics and is also relevant for searches and measurements."
- removed "new"
- We are measuring the production cross section of a process containing two b hadrons, changed "dynamics" to "jet kinematics"
- The goal was to have all of the Wbb and TTbar selections in the same place, as well as to avoid repeating them - this section is just about the samples used and all we want to indicate is that these are the relevant samples to use for our phase space
- Technically this is the one sigma bounds on the input variation used in the fit
- These numbers correspond to the simulation after applying the corrections from the ttbar fits, but before performing the final Wbb fit
- You are correct in your understanding of what the signal strength is, changed "required" to "predicted"
Thomas Ferbel, July 26
Hi Authors et al,
I find this paper somewhat disappointing for several reasons. First of all, the hopes raised in the introduction of tests of the SM and H->bbbar couplings are forgotten by the time you reach the summary section: Did this paper reduce the limits on non-SM contributions? If so, what are these new results?
lets discuss introduction in FR, the goal of the paper is to measure the Wbb cross section that is important in Higgs and BSM searches
Second, in my view, the writing requires some improvement to get ready for the FR. I am surprised It has quite a few violations of our rules, and should be checked again. I have made recommendations for removing the grubby word "shape" and for using more appropriate English words in its place, such as "distribution" or "form," as indicated in my corrections in the attached Adobe Acrobat file.
tried to improve
Connected with this, is the abuse of "QCD" in its usage for "mulitjet" events. Using "QCD multijet" is only correct when it refers to specific MC calculations rather than data from control regions.
this was one of the request during CWR to use this wording
Also, I assume that "variation" means "change," as there is no variation of the kind implied, but only changes made.
we agree and suggest to discuss a possible modification in FR
The structure of the title, abstract, introduction, and summary is not inspiring. Besides the writing (and the repeated summary), there is the change between abstract and introduction in the issue of the associations in the cross section. It is unclear what you wish to impress upon the reader. I would ask you to improve the overall English, and, wth help from your LE, minimize the violations of our beloved CMS PubComm rules.
https://hypernews.cern.ch/HyperNews/CMS/get/AUX/2016/07/26/15:40:57-54421-SMP-14-020-paper-v9.pdf
Kajari Mazumdar, July 18
This is a nicely written paper. I have few comments only.
Type A
l81: add after ".. massive vector boson and jets" --> (V+jets, where V = W
or Z)
- changed
- changed ( "using the Z2* tune")
- ok then
- changed to "... the PDF set CTEQ6L is used and interfaced with PYTHIA6 using the Z2* tune"
Paper Figures
CWR Comments (on Paper v5)
Vassili Kachanov:
The paper is interesting from the point of view of the background estimations in the the Higgs production processes' measurements. This is a new necessary step in the increase of the sensitivity of the measurements to the new physics. This analysis is at 8 TeV and it is a continuation of the measurements at lower energy 7 TeV. In this one both muon and electron decays’ channels were used. The paper describes analysis in detail with many tables and graphs as the illustrative material. The results agree with the previous measurements at 7 TeV and are the first ones at the energy 8 TeV. It is proposed to publish this paper.- Thanks for the support.
Adrian Perineau
Type A: English/Style/Formatting (including figures)
Title: “with two b jets”: the jargon b jets is explained only in the text l.13 Abstract: “two b-tagged jets”: it would be good to be consistent with respect to the title- removed "-tagged" from abstract
- We use the symbols Z and V both in this text, so both are defined here.
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed to "with the factor 0.84+-0.03 calculated in the 7 TeV Z+b analysis [4]"
Type B: Everything else (e.g. strategy, paper structure, emphasis, additions/subtractions, etc).
l 6: the combination of the Higgs mass measurements from ATLAS and CMS is already published and the combined value with its corresponding reference should be quoted- reference added
- changed
- changed to: "to ensure that the triggers are fully efficient"
- changed - definition moved up and removed later
- We did not optimize further than the EG POG recommendation.
- removed "about"
- removed 4F/5F discussion and added :
- changed 5.0 to 4.7
- They are uncorrelated and the text now reflects this. Below is a distribution of transverse mass using QCD MC with inverted and standard isolation cuts.
l 222 “efficiency correction factors” > “corrected efficiency” otherwise is not a cross-section - changed ao "acceptance and efficiency"
- This is result of the measurement, and should be discussed by theorists.
- removed
- reference has been updated is no longer used (ttbar xc)
Albert De Roeck
General comments
On your comment on line 276 on the 7 TeV data: I noticed that the SM prediction used at 7 TeV is from MCFM and it gave 0.55 pb at 7 TeV. Ie the SM cross section at 7 TeV is LARGER than what you give here at 8 TeV (0.50 pb) So the statement is a bit of a cheat since apparently the SM cross section changed quite a bit since the 7 TeV analysis What would be the prediction of the
presently used models/calculations, when ran for the 7 TeV? I guess something
in the line of 0.45 pb, while the measurement was 0.53 pb. Still only one sigma or so too
high but the same tendency as the data in this analysis.
- There are a few differences with respect to the 7 TeV analysis, in particular, the gen jet definiton at 7 TeV included neutrinos while that 8 TeV does not. At 8 TeV, the lepton pT is 30 GeV (as oppsed to 25 at 7 TeV) and at 8 TeV there is no cut on mT.
- We updated our ttbar reference from the CMS measurement to the CMS/ATLAS combination which brings down the uncertainty. We also moved from taking the uncertainty from the fit as 100% of the fit to the uncertainty from the fit. The comparison of the old results and results after these tighenings is shown at the top of the twiki as 0.64 +- 0.03 stat +- 0.10 syst +- 0.06 heo +- 0.02 lumi pb
- We only measured the W+bb cross section
Details
- line 48: what about neutrals in the pile-up? We do not correct for these on some event level based observable? If we do you should mention it here.- These are the standard CHS jets, and are pileup corrected. Neutrals are taken into account via the factor of 0.5
- Neutrals are taken into account via their ratio with charged.
- added "in the signal region phase space"
- changed
- removed 4F/5F discussion and added :
- Nothing too interesting, just for the sake of consistency. We are using the ATLAS/CMS recommended common cross sections listed here: https://twiki.cern.ch/twiki/bin/view/LHCPhysics/SingleTopRefXsec
- updated reference and added value to paper
- changed
- these are the weights being fit for in Steps 1/2 described more fully in the last paragraph of this section
- These are taken from the uncertainties on the cross sections for the given background and text has been updated
- changed to: The same is true for the $\Wbb$ sample, which at LO has only two jets in the final state and therefore is not affected by the jet veto requirement.
- changed
) - added: "following the PDF4LHC prescription" \cite{Botje:2011sn}
- We do not know of any other analyses that have seen this and did not give a dedicated presentation to the JetMET group (though we gave multiple presentations to the BTV POG). A shift of 1-2 sigma is expected from time to time, and we mention in the paper now that 1.3 sigma corresponds to 3.4% in ttbar normalization.
- yes < 1%
- changed
- references added and clarified, (NNPDF=48, MSTW=43, CT=47)
- reference added
- We thought that this was an interesting distribution and may help theorists understand better what is going on.
- reference has been updated
Artur Apresyan
Line 77: not clear yet what the "signal enhanced dataset" means, the cuts are only defined starting Line 121. Move the definition of signal sample here.- We want to separate the discussion of the MC generators used from the statement of signal region selections so have them in different sections. Added internal reference to section
- changed to use words instead of symbols
- changed
- added: "In the multilepton region, the m_T variable is calculated with respect to the lepton named in the channel."
- We are using a data-driven technique to estimate the JES and b-tag efficiency, and the phase spaces used to derive these corrections are chosen to be complimentary: The signal region has a jet veto and a lepton veto, the TT-multijet has the jet veto relaxed and the TT-multilepton has the lepton veto relaxed, so everything is held fixed across the phase spaces with exception of a single final state object.
- We have followed another commenters advice and combined Figures 1,2,3
- We are using the mT variable in three different phase spaces and performing sequential fits.
- Technically we derive a factor which is shape dependent, but in practice, this shape is effectively identical to the original shape, leaving us with normalization changes only.
- We have gotten agreement from the BTV that this is likely a b-tagging issue by observing how our agreement varies with number and tightness of b-tags https://indico.cern.ch/event/436169/contributions/1921818/attachments/1134376/1622924/tmperry_2015july30_TT.pdf and seeing that we observe the same level of agreement as the published ttbar measurements when we use their cuts, and gain a 5% disagreement per tight b-tag in our phase space (25 GeV = relatively low jet pT).
- Assuming that you mean multilepton control sample. Why JES: The difference btw this phase space and the multijet is the lepton veto (we see that MES and EES don't have much of an effect) and the jet veto. The JES shifts jets of ~25 GEV into/out of acceptance and amplifies the effect of the JES.
- We are using the mT variable in three different phase spaces and performing sequential fits and have updated the text to be more clear.
- The shapes of the Wbb and ttbar samples in the mT variable in the signal region phase space have some discrimination power
L181: why do you take a 50% uncertainty on the QCD? where does that estimate come from? Is it just some random "large" number? - We take an uncertainty large enough to allow the QCD normalization to float freely. We have seen that taking larger uncertainties does not affect the final result.
- The important point is that we are using a data-driven technique to estimate the JES and b-tag efficiency with phase spaces that are identical to the signal region except for one specific modification in order to isolate an effect. In this way, effects from things like ISR and FSR cancel.
- 1.3sigma_JES was the rescale factor calculated so these statements are equivalent, but we have moved to using the uncertainty from the fit itself and have changed the text accordingly.
- From the fit, we see that a rescaling is necessary and are attributing it to JES. This phase space is pure TTbar, and we know the bounds of the JES uncertainty. Because the JES interacts differently with different samples, we express the outcome of the fit in terms of how many sigma_JES it corresponds to. Then we rescale the samples in the W+bb region based on this number of sigma_JES.
- changed
- Figure 5 has been updated.
Louis Lyons
GENERAL:
It is a bit difficult to follow in detail your extraction of scale factors. For example, it is not clear why the 2 different final states each are sensitive to just one of the scale factors.- The two ttbar final states are chosen to be idential to the signal region, with the exception of one change: the relaxation of either the lepton veto or the jet veto. It may be more clear to work backwards. In step 2, the lepton veto is relaxed bu thte jet veto is in place. The effect of relaxing the lepton veto is simply to ensure a pure ttbar sample since few other processes contribute to a 2 os lepton 2 jet final state. TTbar is often a multijet process so the for jets around 25 GeV, the jet veto selects different events for different choices of jet energy scale. In addition, both of the jets which are present are b-tagged so the b-tagging efficiency must be correctly modeled. In the details of this analysis, we see that while we technically allow for the possibility of the JES and b-tagging effeciency to have shape dependencies, in practice, the shapes of the shifted and unshifted distributions are not meaningfully different. We therefore must separate the effects of the b-tagging effeciency from the effects of the JES choice. We acheive this in step 1 where we have relaxed the jet veto but kept the lepton veto. We see that the effect of the lepton energy scale is 2-3% and have independent justification to believe that the b-tag efficiency is not properly modeled in this phase space. In a presentation to the BTV, they accepted our summarized the results, that we see a discrepency of ~5% per b-tag in our phase space (but better agreement in the phase space of the published ttbar measurement) https://indico.cern.ch/event/436169/contributions/1921818/attachments/1134376/1622924/tmperry_2015july30_TT.pdf . These b-tag rescaling numbers are therefore what we take from the result of step 1 and apply before moving to step 2. In the step 2 fit, when we allow the b-tag rescaling to float with a 1 sigma uncertainty large enough to return to it's original value, we see that it stays nicely at 1. Additionally after rescaling by the JES from step 2, we perform a closure test by refitting in the multijet region and again see that the scaling does not move.
- Do you mean the 7 TeV, since this is the only one Wbb at 8 TeV? In general you are right that adding electron channel reduces the uncertainties very little, but it is another measurement and it effects also the final cross section value
SPECIFIC POINTS:
Line 46: Which sum?- Text has been clarified
- removed 4F/5F discussion and added :
- If the factor is 1+epsilon, then the uncertainty was taken to be epsilon. We are now taking the uncertainty from the fit.
- This number was a result of putting in the 100% uncertainty on the fit and is no longer in place.
- No, this is the gaussian 1 sigma bound. Text clarified
- removed
- We have clarified that in the multilepton region, the lepton being named is the one used in calculating mT
- This was a conservative estimate, including both the JES and the rescaling of the JES. Now we are taking just the uncertainty from the fit (Step 2).
- Yes, for these other variables, the QCD shape is taken from the low mT sideband and have been validated using QCD MC. Illustrated here are comparisons above and below mT=30 GeV for QCD MC in the dR(bb) and pT_Lep variables.
Table 2
a) The numbers in the 'muon, initial' column do not add up to the data number. - This is one of the motivations for performing a fit.
- Not worrying, but this is what we see.
- Yes, the signal strength is estimated higher for electrons than muons. The acceptance x efficiency are not actually used in the calculation of the cross section, and is affected by the rescalings in the fits. The numbers quoted in the text showed higher precision than we should have (+-10) and we have modified the text
- This is what we observe with the method we described.
- changed
- This is because of our use of no neutrinos in the gen jets at 8 TeV while at 7 TeV the gen jets had neutrinos in them.
- Yes, they are both needed.
Greg Landsberg
PHYSICS COMMENTS:
L7: add the CMS long discovery paper as an extra reference.- added reference
- section added
- added
- Yes we use Beta Star for both leptons
- removed equation
- changed
- references added
- changed to all words
- added
- added
- sentence (and paragraph) has been rephrased and clarified
- removed
- moved
- This was a result of the inflating up to 100%, we now take the uncertainty directly from the fit.
- added "the uncertainty on the rescaling sets the one standard deviation bound on the rescale factor in subsequent fits." and the text has been clarified in the description of the procedure
- changed
- you are correct, changed
- added (FEWZ 3.1)
- defined
- We start with the standard recommendation and then do the 3-step fitting procedure described to rescale it. This uncertainty is the uncertainty on the final fit (which is equivalant now to the output uncertainty of the b-tag efficiency fit in step 1) .
- MCFM7.0
- new Figure 5 has been made
- added
STYLE COMMENTS:
Title: suggest "Measurement of the production cross section of the W boson in association with two b jets ..."- changed
Abstract:
L7: fiducial region defined by the transverse momentum- abstract changed
- abstract changed
Introduction:
L3: at the CERN LHC is relevant- added
- changed
- changed
- changed
- changed
Event selection and reconstruction:
L51: Missing transverse momentum, p⃗ missT, is defined;- kept as MET
- changed
- changed
- symbol kept
- We were previously asked to change to follow reference who uses k_t in the title.
- We were previously asked to hyphenate all "b-tag"
- changed
- not changed
- changed
Simulated samples:
L77: After all the requirements for the signal-enhanced dataset are applied;- changed
- changed
- changed
- changed "or" to "and"
- We were previously asked to make it k_t
- changed
- changed
- kept version number
- kept version number
- kept version number
- kept version number
- udated reference cross section and citations
- kept version number
- changed
Analysis strategy:
LL115-116: Since initial shapes and normalizations of all contributions, except for QCD, in the fit are taken from simulations, it is;- changed
- changed
- changed to "particle list level"
- changed to "bottom"
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- consistent style to keep b-tag hyphenated
- consistent style to keep b-tag hyphenated
Systematic uncertainties:
L174: effect on the signal yield.- changed
- changed
- changed
- changed
- consistent style to keep b-tag hyphenated
Results:
L190: the b tagging efficiency;- consistent style to keep b-tag hyphenated
- description changed
- merged Figures 1,2,3
- changed
- changed
- changed
- changed
- changed
- changed to b-tag eff rescaling
- changed
- changed
- changed
- changed
- changed
- changed
- changed to "b-tag eff rescaling"
- changed
- changed
- changed
- changed to "after the fit"
- changed
- changed
- changed
- merged with Figure 1
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed to "b-tagging efficiency"
- changed
- changed
- changed
- changed
- changed
- changed
- kept because both Pythia6 and Pythia8 give similar DPS numbers
- changed
- changed
Conclusions:
L285: pℓT>30 GeV, $|\eta^\ell| < 2.1;- taken from abstract
References:
Refs. [1,2,6,8,10,11,21,29,36-39,41,44,48]: separate parts of journal name with a space: e.g., Phys. Rev. Lett.- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- reference removed
- Latex takes care of this automatically
- changed
- changed
Avto Kharchilava
Type A
Eq. (1): Instead of sum(pT^gamma)+sum(ET^neutral), use sum(ET^neutral), or sum(pT^gamma)+sum(ET^neutral hadron)- equation not changed, updated explanation
- changed
- This is beta star
- changed
- changed
- added
- TT-Multijet Prefit
- TT-Multijet Fitted
- TT-MultiLepton Prefit
- TT-Multilepton Fitted
- Wbb Prefit
- Wbb Fitted
L130-131 seem inconsistent with L86-87. In L86-87, the shape of the Wbb is taken from dedicated W+bb sample, while L130-131 says the shape is taken from W+jets samples. - Has been clarified in text
- The shapes are different enough to have some discrimination power as can be seen in a direct shape comparison here:
Paragraph at L130 describing jet flavor definition in simulation:
"Jets with distance parameter smaller than R=0.5 with respect to lepton are removed from the event." This is done in order not to count prompt leptons as genjets. So leptons from b-hadrons should not be considered in the procedure otherwise it will remove jets with semileptonic b decays. This indeed is stated in the sentence "Generated leptons originating in simulation from the decay of b-hadrons or tau-leptons are not considered." However, it's unclear why leptons from tau decays should not be considered in the procedure. Does it mean it's ok if they get reconstructed as genjets? Why don't simply state that "Jets with distance parameter smaller than R=0.5 with respect to prompt leptons from V decays are removed from the event." - changed
- clarified text on systematics
- yes they did
- added
- We now take this to be a closure test and the uncertainty is now taken directly from the uncertainty on the multijet fit.
- We no longer adjust the uncertainty on the fit and take it from the fit itself, so 3.8% in this case.
- The k-factor does not have a shape dependence so does not affect the fit. If it were higher, then we would measure a correspondingly lower signal strength, resulting in the same measured cross section.
- We mean at the (neutrinoless) jet level. For MCFM, this is accomplished by multiplying the particle level estimation by a hadronization factor.
Type B
Abstract: Latex: 19.8 fb-1 => 19.8~fb-1- changed
- changed
- changed
- changed
- added
- changed
- followed another comment
- changed
- changed
- defined now
- changed
- changed
- changed
- changed
- removed
- taken from abstract
Figures
Could improve on the fill pattern: W+bb and W+cc have ~same color- changed W+cc color
- changed to "Multijet" in legends
- changed
- remade Figure 5
- remade Figure 5
- included description of four- and five-flavour schemes and updated plot/caption
Tables
Add outer borderlines.- The pubcomm guidelines recommend against this: https://twiki.cern.ch/twiki/bin/viewauth/CMS/Internal/PubGuidelines#Figures_and_tables (last bullet)
- The reference has been updated and the uncertainty on the cross section has gone down.
- done
- The final normalization of the sample comes from a change in cross section (signal strength) as well as from the effects of systematic uncertainties which may be correlated across different samples.
- make plots and tables consistent
Salvatore Rappoccio
Lines 87-89: Does this procedure accurately pick up the components of W+bb from flavor creation and gluon splitting? Long ago this was checked and found to be lacking. Are your phase space cuts sufficient to suppress the gluon splitting component? It would still be good to quantify this.- How does flavor creation make b quarks? The CKM matrix makes this a very small effect. Gluon splitting is what we are looking at.
- Flavor creation is negligible and gluon splitting is what is being studied.
- It is valid for gluon splitting.
- Flavor creation does not play a role here.
Theodoros Geralis
Type A comments (English/Style/Formatting)
Abstract: “ … a fiducial region having transverse momentum …” propose to write “ … a fiducial region defined by leptonic transverse momentum …” L 85: the numbers 20 and 30 GeV are given without a reference or an explanation.- numbers removed
- fixed
- Now described in words
- changed
- changed
- all simulated samples
- changed
- changed
- changed
- now defined
Type B Comments (Strategy, paper structure, … )
Title: The title as it is implies an inclusive cross-section, while W decays to (e,nu) and (mu,nu) are considered only Abstract: The same remark for the first phrase of the abstract. * We think this is clear from the abstract, but this can be discussed during the final reading L 18-20: ambigious meaning, it could also be interpreted as that what follows extends the data collected at 8 TeV.- changed to "using" to "and uses"
- removed, this is covered by "jet kinematics"
- changed
- clarified
- added "processes" and changed "prove" to "verify"
- W+b is CKM suppressed while W+c is not. So W+bb and W+cc are more simlar (gluon splitting) than W+cc and W+c.
- Yes. Below are the isolated and non-isolated regions using QCD MC
- And for the other variables, we alse see good agreement with our choice of QCD sideband (mT<30 GeV)
L 150-153: Is the QCD normalization extracted in the MT<20 GeV region the same as the one you get when you use the anti-isolated region? - yes
- This is the ttbar multijet region (third jet requirement) and includes mT>0 . The QCD is allowed to float in this fit.
- Even with the 50% uncertainty on the fit, we see that the QCD contributes 2-4% to the overall uncertainty in the measurement
- That is correct, the scale and PDF are taken (as well as luminosity) separate from the fit. The propagation of the fit (3rd fit) on the JES and b-tagging is reflected in the last column of the Table 1 which indicates that the effect of the given systematic on the final quoted uncertainty.
- We are now taking the uncertainty directly from the fit.
- clarified text
- clarified text
- JES is allowed to fit with gaussian limits, the 1 sigma bounds of which are set by the fit in Step 2.
- No we do not
Tristan Du Pree
The main comments are on the fitting procedure and its description, and some unclearity about the main background: ttbar. There are also some comments about the background uncertainties (JES and BTV) and the used generators. General analysis comments are given first, more detailed stylistic comments are below. = = = General analysis comments = = = General comment: It is surprising that the analysis ends up with a ~20% uncertainty of which a large fraction (13.0%) comes from b-tagging. It is expected that the b-tag data-MC scale factors have much smaller per-jet uncertainties in the relevant pT range. Now it appears that this analysis instead takes larger uncertainties from a control region, but there is not sufficient details in the paper draft to understand the reason behind. (why is it needed at all? wouldn't it be proving the BTV results wrong if it's serious (but even this is impossible to judge from the paper)? and similarly the size of the JEC uncertainty used by all other run-1 analyses? why not perform a joint fit of the three regions? why can we measure the ttbar cross section with 4-5% precision but have to put a 7.4% uncertainty here?)- The procedure is needed because we see that the b-tag efficiency scale factors do not describe the data in our signal region phase space. It does not prove the BTV results wrong because we showed good agreement in the ttbar phase space measured by CMS. The JES uncertainty is also within the recommendation and our analysis is particularly sensitive to JES because of the jet veto. We do not perform a joint fit because it is less stable. We have updated the ttbar reference and use a lower uncertainty
- Yes it converges in closure tests at all steps.
- They do not move from their new central values
- The uncertainty on the JES is now taken from the fit
- It is the envelope, and we have added the corresponding percent normalization difference in the ttbar sample (3.3%)
- done
- They go in opposite directions which is why the approach of doing a simultaneous fit does not work.
- We see no impact from top pt reweighting which is expected given that we see no shape difference even when varying the JES. The hadronization correction is reliable and we do not have a Powheg sample
- We have updated the ttbar cross section reference ( 239 from CMS to 241.5 from CMS/ATLAS) which is the joint CMS/ATLAS result and has a lower uncertainty
- We see good agreement with the experimental value when using loose b tags. It is only when we move to our phase space that the disagreements mount, as was presented to the BTV POG and accepted as being related to the b tagging efficiency
- Yes, they are compatible, but with a central value that better describes our phase space
- This is along the lines of the first approach this analysis tried, to do simultaneous fits. The results are consistent within uncertainties, but less stable.
- yes we have applied the SFs
- Taking advantage of the pT dependence of the scale factors already derived by the BTV, we found that we only need to adjust this relative normalization.
- text has been clarified
- We have found that we need to separate the fit to isolate the effects of b-tagging and JES. We now take the uncertainty directly from our fits instead of inflating them to 100\%
- noted
- The text has been clarified. The JES is shifted and new templates are made and these are used in the fit. The scale factors are consistant with the official b-tag and JES factors. We are confident that this is coming from the b tag as discussed with the BTV POG, and the JES is expected (and also observed) because of our jet veto. Yes, this discrepancy has been discussed with the BTV POG.
- b-tagging has an impact on the shape of distributions in principle, but in practice in this phase space, it only affects the normalizations.
- Yes, we start the Results section with a quick overview of the procedure
- This was a result of our inflation to 100% and is no longer the case
- We see that the variation of renormalization and factorization scales give a final uncertainty of 10% and top pT reweighting has no effect
- we compared aMCatNLO and Madgraph (+pythia) obtaining similar results for the hadronization factor using WBB samples
- The problem is not pythia8, but the hard interaction. the 4F sample produces directly at hard level a W+ bb sample. The contribution from events with WMuNu + another interaction that gives the bb par, cannot be done simply with pythia. Yes, the subsequent MPI interactions (more than 2 b) are included, but they are negligible in comparison
- Why is this a lot - do have a reference showing it should be less in mind that you could give to us? It is a value that is cross checked both using Madgraph samples (WMunu inclusive + bbar production, explicitly done with DPS) compared to an older estimation assuming fiducial XSec_WMuNu x fiducial XSec_BB / sigma_Effective.
- Figure 5 has been updated
- We used the samples that were available
- We are instead using the ATLAS/CMS recommended single top cross sections: https://twiki.cern.ch/twiki/bin/view/LHCPhysics/SingleTopRefXsec
- The text has been reworded - "For the signal distributions, the \Wbb component of an inclusive \Wjets sample is used, with the shapes of the distributions taken from a dedicated high-statistics generated sample of exclusive \Wbb."
- We see the slope and thought that others might find this distribution interesting.
- Statement removed. We will leave it to the reader to decide if they would call this agreement or disagreement.
- changed
- changed
- first instance refers to the process itself, second to our measurement
- changed
- added "and measurements"
- changed
- changed
- changed
- There are many searches ongoing in CMS in this general category and we did not want to mention a specific one at the exclusion of others.
- changed
- changed
- changed
- added reference to LHC
- changed
- removed
- We expect that most readers will be familiar with the concept of lepton isolation so it was worth mentioning specifically that the triggers have some isolation requirement. As you point out, isolation is described more fully a few lines below.
- changed "triggers with a loosely..." to "triggers which require a loosely..."
- changed
- changed
- changed
- Explanation clarified
- line removed
- changed to "radius parameter"
- changed
- changed
- changed
- They do not, they are the CSV Tight scale factors.
- changed
- changed
- changed
- changed
- line has been dropped
- same as V+jets, added
- changed
- removed
- removed
- removed
- removed
- removed
- changed
- changed
- changed
- CTEQ6M, added
- changed
- reference dropped
- changed
- changed
- changed
- We want to state all of the selection cuts concisely in one place.
- changed
- We do not make this requirement. With the selection as is, the region is already pure ttbar.
- description moved
- W+c is included in the category (W+cc is not).
- changed
- W+b is CKM suppressed while W+c is not.
- changed
- Yes this includes the lepton from the W.
- W+c happens while W+b doesn't.
- How about where it is then?
- changed
- description changed
- changed
- changed
- changed
- This is covered by the 50% systematic uncertainty and was determined by looking at Isolation vs. MT. We chose our QCD isolation sideband by inverting the cut at the loose isolation threshold.
- l.158: "on the third" -> "on a third" - changed
- changed
- changed
- at leading order W+bb has exactly two jets
- The shapes are too similar for the fit to distinguish between the systematics - both uncertainties are found to have essentially no effect on the shapes of the distributions - only normalization. We therefore use the three step process to isolate the different effects. Note that this is a different statement than that our choice of variable to be fit does not discriminate between the signal and backgrounds.
- changed
- changed
- changed
- description updated
- changed
- description changed
- changed
- The b-tag efficiency is considered as the main source of data/MC disagreement, though the other uncertainties are applied.
- The b-tag is fixed here, and we confirm with a closure test on the post fit result (applying it in both the tt-multijet and tt-multilepton regions) allowing it to float. The result of these closure tests is unity.
- changed
- changed
- changed
- changed
- changed
- removed
- defined
- reference added
- changed
- changed
- changed
- removed
- line removed, but you are correct
- procedure changed, now taking uncertainty directly from fit
- changed
- changed to reweighted
- We show all of the independent samples which are allowed to float in the fitting procedure.
- changed
- Correct, and there is also the difference in the trigger required.
- changed
- changed
- changed
- changed
- One was for the JES uncertainty and one was for the uncertainty on the rescaling. We now are just taking the uncertainty from the fit and the table has been updated accordingly
- These are the b tag efficiency corrections provided by the POG and used as input in Step 1.
- changed source of ttbar cross section, and updated uncertainty
- same uncertainties with the exception of the rescale factors that haven't been fit for yet. Luminosity is added separately.
- changed
- changed
- changed
- correlation across all simulated samples
- changed
- changed
- Largely the same events, but not exactly, and mT is being calculated via a different lepton.
- W+c and W+cc are two kinematically different processes and we would like to distinguish them in principle since we have some W+udscg contribution but essentially no W+cc in the signal region.
- changed
- merged with Fig 1
- merged with Fig 1
- Do not understand this comment. Is the suggested sentence this? "...are also produced by applying the results from should be simultaneous to the simulated samples"
- changed
- changed
- changed
- changed
- changed
- line removed
- added total MC to bottom
- The W+udscg normalization changes on the order of 5\%
- changed
- changed
- correct
- changed to "total uncertainty of the fit"
- changed
- changed
- Updated plot colors
- changed
- changed
- changed
- changed
- changed
- text has been updated
- The W+jets sample is used and the cross section taken from there - but the normalization of the Wbb should not matter. A higher initial cross section would lead to a smaller signal strength and the final measured cross section would be the same.
- changed
- changed
- changed
- we varied all listed here, but HERA was not the largest anyways, so does not set the bound on the uncertainty
- correct, added "simultaneously"
- Electron channel has gamma+jets and electrons aren't as clean as muons.
- changed
- changed
- Showing that two schemes give the same result is something we consider "important feedback".
- sentence changed
- changed
- changed to just "simulation"
- changed
- Kept for clarity
- sentence removed
- Figure and caption updated
- Figure and caption updated
- Figure and caption updated
- Figure and caption updated
- changed
- now the same as in absrtact
- now the same as abstract
- all references updated
- changed
- reference taken from inspire
- changed
- removed
- changed
Andrea Rizzi
Type B
Abstract: the last sentence is excessively long, with a subject five lines (!) before the verb ! Break it, for instance as “The W+bb production cross section is measured in a fiducial region having lepton transverse momentum, Ptl> xx and eta<yy, with exactly … . The measured cross section is found to be sigma= zzz in agreement with standard model predictions.” Note that you do not need to define \ell because you have already specified that the measurement is done with muons and electrons.- updated abstract
- Good point, changed
- changed
- moved
- added
- reference 8 is about 7 TeV and this analysis is about 8. Sentence has been reworded.
- removed
- The charm and light rate are from the BTV and the b rate we measured to be 40% instead of 50%. We have made many presentations to the BTV group on the topic of b-tag efficiency, for an example and link to more presentations, see: https://indico.cern.ch/event/436169/contributions/1921818/attachments/1134376/1622924/tmperry_2015july30_TT.pdf
- reworded
- changed to be stated in words
- added description
- The uncertainty associated with the jet veto comes from the uncertainty associated with the choice of jet energy scale and this is what is being fit for in step 2.
- Are we talking about 2 real bs in one jet, and a second light jet mistagged? Because we require both jets to be CSV Tight-ly tagged, this makes a negligible contribution to the signal region and we do not include a dedicated systematic since it is covered by the btagging uncertainty.
- The pure process W+(single b) is CKM suppressed (much more so than W+(single c) is) and does not make a contribution to our signal region - especially since we require both jets to be tagged (CSV Tight)
- We consider W+c as part of W+light and W+cc in its own category. This is motivated by the difference in production mechanism between W+c and W+cc.
- It does not matter to the analysis if they end up in W+light or W+cc since they constitute a negligible fraction of events.
- It is adjusted in the standard isolation sample. Text has been clarified
- text has been modified
- Text has been clarified.
- Statement has been removed
- This was a result of our 100% uncertinty. We now take the uncertainty directly from the fit
- It is now mentioned in the text that this corresponds to 3.4% for ttbar.
- The JES affects the different samples to different extents so the fit in the ttbar is first converted from a change in normalization (normalization only because we have seen that the shapes are identical, and we do include them in principle) into the number of sigma_JES this corresponds to for ttbar. The other samples are then shifted relative to their own JES.
- This is a description of other variables than mT being shown in the postfit signal region.
- The normaiozation is taken from the final output of the step 3 fit (as are the normalizations for everything else). It is only the shape that is taken from the low mT sideband. Validation plots of the mT<30 sideband using QCD MC are shown below
lines 228-229 11% cannot result from 80% times 40% … what you
probably have to clarify is that there are these two contributions,
but also the geometrical acceptance. - It is 40% per jet and we have two jets so 0.40*0.40*0.80
- When we have to separately measure the DPS, we measure it with a large uncertainty, and when it is already included in the sample, the uncertainty is smaller. In both cases, it is important to include.
- The statement has been removed
Type A
- General: the draft is full of typos, please pass it through a spell-checker, e.g. - line 87 inclusive - line 105 collected - line 165 region - line 166 efficiency- changed
- changed
- changed
- changed to particle list level
- changed
- Figure 5 has been remade. 2012 was referring to the dataset used but has been removed
- mu is already being used for renormalization and factorization scales
- changed
- changed
- reference added
Sijin Qian
Type B (physics)
noneType A (style)
In general (1) Throughout the paper (including in Abstract, in Figure's and Table's captions, in Tables, etc.), to be consistent with the good examples in this paper (e.g. on L255 and Table 2 (the header column and the 8th row under the header row), etc.) the spaces before and after the symbol "+" in all the expressions of "xxx + jet", "W + bbbar" and "W + ccbar", etc. should be removed, e.g. L19: "W + bbbar" --> "W+bbbar"- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- have been advised to remove the line entirely
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- kept abbreviation
- changed
- changed
- this discussion has been removed
- not changed - "I see two houses of different color" vs. "I see two houses of different colors" the second implies that the two house colors are different than something else
- changed
- changed
- changed
- changed
- changed
- changed to Transverse mass [GeV]
- changed to dR(b,b) and Lepton transverse momentum
- changed
- changed
===========================================================================
normalization |
| uncorrelated
|=============
| norm. + shape | correlated
========================
luminosity 2.6%
theory (scale+PDF) " -->
" | Uncertainty | Variation | Effect on the measured cross section
===========================================================================
Normalization |
| Uncorrelated
|=============
| Norm. + shape | Correlated
========================
Luminosity 2.6%
Theory (scale+PDF) " - changed
- changed
- Table 1 now appears before Figure 1
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed
- changed to DPI (S was for scattering)
- changed
- Figure and caption have been changed.
- conclusion now matches abstract
- added
- changed
- changed
- changed
- changed
- changed
- added
- changed
- changed
- changed
- citation removed
- changed
- changed
LE Comments (on Paper v3) Don Lincoln
Basic questions:
In the abstract, it talks about the cross-section pp -> W(l nu) + bbbar). It is unclear to me if the cross section is really pp -> W + bbbar or W(l nu) + bbbar. In short, given that the W -> e or mu 2/9 of the time, is the 2/9 in there or not? This is just a language thing. Are you correcting for the branching fraction? Or is it truly a cross-section for the lepton and electron final state? We are not correcting for the branching fraction, but are performing the measurement in the W(e,nu) and W(mu,nu) channels. In the abstract, you say “with exactly two b-tagged jets.” It is unclear if that means “exactly two jets, both of which are b-tagged” or whether it means “exactly two b-tagged jets and no other jets.” Please clarify. Changed From : ... exactly two b-tagged jets having ... Changed To : ... exactly two b-tagged jets with $p_{\mathrm{T}}>25$ GeV and $|\eta|<2.4$ and no other jets with $p_{\mathrm{T}}>25$ GeV and $|\eta|<5.0$ ... Is there a |Z_vertex| cut? There is no such cut. There is a cut ensuring that the the leptons come from the primary vertex, designated as the one with the highest sum pT of the tracks. L87 – 89: You talk about the four flavor scheme as compared to the five flavor scheme. This is pretty confusing, given that bottom quarks are obviously the 5th flavor. I think this section needs a little massaging to clarify the ideas. Had: The $\Vjets$ samples are generated using the five-flavor scheme that includes massless b quarks in the initial state. Additionally, a signal $\Wbb$ sample is generated in the four-flavor scheme which has more statistics than the $\Wbb$ component of the $\Wjets$ sample generated in the five-flavor scheme. Therefore, in this analysis the four-flavor sample is used for the signal shape contribution, with a normalization taken from the $\Wbb$ component of the five-flavor $\Wjets$ sample. Became: For the signal distribution, two generated samples are used. The normalization of the distribution is taken from the $\Wbb$ component of the $\Wjets$ sample and the shape of the distribution is taken from a dedicated $\Wbb$ sample. The reason for this separation is that the $\Vjets$ samples were produced using the five-flavor scheme in which the b-quark is treated as a massless parton, while the dedicated $\Wbb$ sample was produced using the four-flavor scheme in which the b-quark is treated as a massive particle which decouples from the evolution equations. The five-flavor scheme is more accurate than the four-flavor scheme in the limit $\mu_F >> m_b$, but the four-flavor sample was generated with more events passing the signal region phase space selections, and hence has a smoother distribution in the variables used in fitting and cross section extraction. * We could also mention that cross checks were performed using exclusively 5F and exclusively 4F shape/normalization and the results were compatible with what we show. L127 – 139: I think that the selection cuts in this passage are at MC truth level, but I am not 100% sure. Yes, the splitting of W+jets into W+bb, W+cc, and W+udscg is done using MC truth information. Removed: Exactly two b jets with $\PT>25\GeV$ and $|\eta|<2.4$ are required in the signal region. I don’t know what the last sentence implies. Those jets are “rejected.” What does “rejected” mean in this context? Is the event rejected? Are the jets removed from the event? "rejected" means "removed from the event" and the text has been changed accordingly Why the asymmetry between the fact that if a bottom quark is created at ME or PS level, it is a W +bbbar event, while two distinct charm jets means W + ccbar, while charm that is not resolvable into jets is W + udsgc? This difference comes from the difference in production mechanism. While W+c can be directly produced via the interaction of a 1st generation quark with a gluon and is detectable by CMS (see figure 1 in http://arxiv.org/pdf/1310.1138v2.pdf), the analogous W+b final state can not and makes a negligible contribution to the signal region. In contrast, W+cc where the cc is produced via gluon splitting has very similar kinematics to the signal region: W+bb. So the lack of distinction between W+b and W+bb at the generator level was chosen so as to be as inclusive as possible while the distinction being made between W+c and W+cc comes from the fact that both final states occur at a nonnegligable rate but arise from different processes.
I found the entire paragraph L 161 – 168 a bit confusing. I also am not sure if the shapes referred to in L 161 is the eff_b or JES shapes or both?
These "shapes" referred to the shapes of the mT variable in the simulated samples (MC and QCD).
Had:
First, the shapes obtained from the simulation together with the QCD shapes are fitted to $\MT$ data distributions in the $\ttbar$-multijet region. The obtained b-tagging efficiency rescaling factors are measured separately in the muon and electron control samples and averaged before being applied. The reweighted simulated samples are used in the next step, where the $\ttbar$-multilepton region is used and the jet energy scale in the simulation is adjusted. As a result of these two steps, the simulation is expected to properly describe the $\ttbar$ contribution and can be used to extract the number of $\Wbb$ events from a fit in the signal region.
Became:
First, using the simulated samples detailed above,
a fit was performed in the $\ttbar$-multijet reigion
using the $\MT$ variable.
The result of this fit gives an estimation of the
b-tagging effeciency rescaling factor, which
is measured separately in the muon and electron
channels and averaged before being applied.
The reweighted samples are then used in the next step
where a fit to the $\MT$ variable in the
$\ttbar$-multilepton region is performed and
the jet energy scale in simulation is adjusted.
As a result of these two steps, the simulation is
expected to properly describe the $\ttbar$ contribution
and the final step is to extract the number of
$\Wbb$ events from a fit in the signal region.
L178 – 184: I don’t understand how putting the uncertainties at 100% allow the fit to remeasure them. Or maybe I do. But I don’t understand what is going on there. Are these backgrounds essentially allowed to float within a 100% uncertainty?
Yes, this is correct. Because we are remeasuring these scale factors whose effects are very difficult to disentangle, our strategy was to adjust the simulation according to the fit, but allow future fits to recover the initial scaling at 1 sigma. If we saw that successive fits pulled the scale factors substantially back toward their original value, this would be evidence of bias in our method, but we see instead that the rescalings stay fixed. This is also confirmed by performing closure tests, re-performing a fit after extracting and applying the scale factors and seeing that it then delivers a scale factor consistent with unity.
And I am curious about the asymmetry between allowing the b-tagging and JES uncertainties to
float between 0 – 200%, while the PDF uncertainties are a factor of two (50 – 200%)? This asymmetry confuses me.
For the b-tagging and JES uncertainties, the uncertainty was chosen specifically with the intention of setting particular values as being one standard deviation away from the central value (namely, the value that "undoes" the rescaling). The uncertainty associated with the change in scale is estimated by changing the renormalization and factorization scales up and down by a factor of two, but the final value is ~1%. The PDF uncertainty is estimated by taking the bounding window of estimations from a variety of PDF sets, and is ~10%, thus giving the combined PDF+scale = 10% given in Table 1
I understand the equation on page 7, give or take. There is an implied “ = “(N^Data_signal / N^MC_signal) x (N^MC_generated / Lumi)” step, or at least so I think. The first term is taken to be the signal strength, while the second term is sigma_gen. I think? I have never heard the term “signal strength” and am confused by it. In addition, It is clear what you are substituting in when you equate the second and third terms. But if I understand that properly...and I probably don’t...N^MC_signal includes acceptance and efficiency, but it also has a background contamination. I guess N^Data_signal also has a background contamination, and it looks like you are assuming they are the same. I think?
You are correct both in the missing equality and in the problem that one encounters when trying to use this equation directly substituting N_events for the different variables, and this is what motivated us to use signal strength in the first place. In performing a fit, the final number of signal events will change as a result of two sources: the effect of uncertainties which are correlated across different samples (things like b-tagging, JES, MES, EES, ...) and the effect of a change to the cross section of the signal sample itself. Factoring the total change in signal normalization into these two components then, the signal strength is just the component coming from the change in cross section, and thus is applied to the generated cross section used to find the measured cross section as indicated in the equation.
L228 & L227. I don’t know how to reconcile the 11 (13)% and the 80% and 40%. (Because isn’t 80% x 40% = 32%?) The missing piece here is that 40% is the btag efficiency per jet, so the calculation is 80% x 40% x 40% = 12.8% which is consistent with the 11/13% quoted above.
To get 11/13% we invert the equation on P7 and the numbers in Table 2 acc x eff = ( n_wbb ) / (lumi * xc * ss) to give mu: 0.1329 ele: 0.1148 We estimate the efficiency of the lepton requirements and (per jet) b-tagging by looking at the raw number of events passing our selections
Added statement "per jet" to Paper
L252 – 253: you say that it agrees with the 7 TeV Z+b analysis. I know you give the reference, but it would help the lazy reader if you were to quantify your statement here.
added
Editorial rules:
In the abstract, you use many symbols, like p_T^l, eta^l, etc. The rules are that the abstract should not include user defined symbols. The question is whether things like p_T count as user defined symbols.I do not think that we have user defined symbols in the abstract. For example, pTlep is given as $p_{\mathrm{T}}^{\ell}$ . If there exists a list of allowed symbols, that would be useful, but I have not found such a list.
Minutia
L 4 – 5. The phrase “a background to standard model Higgs boson production associated with a vector boson, where the Higgs boson decays into a bbbar pair” would be better as “a background to vector boson production in association with standard model Higgs boson production in which the Higgs boson decays into a bbbar pair.”done
L6: I’m not sure that you need to spell out the acronyms of the two experiments. While it’s not wrong, it’s ugly and jarring.
done
L9: Remove “(minimal or not)”
done
L10: through -> using
done
L11: hadrons -> hadron^M
done
L12: add comma: boson,
done
L 13 – 14: replace “to bjets as jets which originate from hadronization of b quarks” with “to jets which originate from the hadronization of b quarks as b jets”
done
L 17: “In the past, the” -> “The”
done
L18: Remove “by the ATLAS and CMS collaborations,”
done
L19: Remove “previously”
done
L 28 & L35. Shouldn’t pseudorapidity be defined?
defined
Equation 1: Why I^{rel} and not just I?
changed to I
L65: that -> which
done
L72: “Tight” is jargon. Drop and rewrite as “both jest pass a threshold which has”
done
L77: You say “signal region,” which seems a little jargon-y to me. Maybe “signal sample”? I’m a little iffy on this suggestion.
changed
Had:
After all signal region selection requirements are applied,
Became:
After all selection requirements for the signal enhanced dataset are applied,
L95: You define diboson here and yet use it on L79. Maybe define it earlier?
defined on L79 now
L100 – 102: “The ttbar...measurements [42].” Is a bit unclear. Maybe “The ttbar cross section was determined from data [42] using ATLAS and CMS data.” ????
how about:
The ttbar cross section was determined from data colleced by the ATLAS and CMS detectors [42].
L113: “fit, except QCD, are taken from simulations” -> “fit are taken from simulations, except QCD.”
done
L145: You say “all other backgrounds.” That means “all backgrounds not yet mentioned,” right?
No, this means all backgrounds other than QCD. We apply the QCD selections to Data and MC, then take the difference as the QCD shape.
L151: affect significantly -> significantly affect
changed L159: you explicitly call out LO and I see how that is relevant in context. But I don’t see where in the background calculation discussion where it was called out earlier. It seems to me that that should be done. Both the 4F and 5F samples are generated with MadGraph as mentioned on L82. No change made - ok?
L161 & L166: It says three steps and two steps and I imagine that this means that the steps are broken up into 1 + (2 + 3), but I can’t see how that was done.
It isn't. This was just a description of steps 1 and 2. Step 3 is the fit in the signal region described in the Results section.
L 161 – 166: I am having trouble understanding the sentence “First, the shapes...multijet region.” Please clarify. Does this fit include the W + bbbar?
Agreed, this was awkward. The intention was to say that everything was fit (but QCD comes from a data driven technique, so I didn't want to call it simulation)
The whole paragraph changed to:
The fit procedure thus consists of three steps.
First, using the simulated samples detailed above,
a fit was performed in the $\ttbar$-multijet reigion
using the $\MT$ variable.
The result of this fit gives an estimation of the
b-tagging effeciency rescaling factor, which
is measured separately in the muon and electron
channels and averaged before being applied.
The reweighted samples are then used in the next step
where a fit to the $\MT$ variable in the
$\ttbar$-multilepton region is performed and
the jet energy scale in simulation is adjusted.
As a result of these two steps, the simulation is
expected to properly describe the $\ttbar$ contribution
and the final step is to extract the number of
$\Wbb$ events from a fit in the signal region.
L173: add comma: cases,
done
L174: I think the “ prior to “norm” should be ``.
changed
L175: add comma: procedure,
done
L191 – 192: I don’t understand in detail the implications of this sentence.
It means that we rescaled the simulation using the scale factors described in the previous sentence. Removed "accordingly".
Table 1, [caption]: what do you think about changing UNC for X? I spent a lot of time wondering about what uncertainty UNC was.
changed
Also, sentence starting in line 4 “In the ‘variation’...UNC in the fitting procedure.” Seems to have an extra word or something that makes it not entirely clear to me.
Not sure what to do - didn't change anything
Table 1: JES rescale line & L196: I’m confused by the 1.3 & 1.6 disparity.
1.3 is correct, changed.
I’m also a little confused by how these tie into the “in the fit” on L194.
Clarified by removing the statement: The energy scale in the fit varies within its uncertainties, $\sigma_{\mathrm{JES}}$.
Figure 2 [caption] vs. figure 2 [figure]: You use different terms, specifically “ttbar- multilepton phase space” vs. “ttbar multilepton control region”. Personally, I think it might be best to use the phrase “ttbar multilepton enhanced data set.”
changed to ttbar multilepton enhanced data set in both places
L213: There is an extraneous “)”
removed
L240: add comma: scheme,
added
Comments on v2
Abstract: try to use the same \bar comment for the anti-b as in the text later (here it appears twice)
- changed
- will wait for LE comment
- changed
- changed
- changed
- changed
- Yes, this is understood and comes from the fact that we are not just averaging the two fit results, but are combining the two channels and performing an entirely separate fit. Before adding the G+J sample which affects only the electron channel, we had previously seen a larger bias, but now the difference from the combined result is +0.03pb (ele) and -0.04pb (mu), so the combined fit is essentially equivalent to the average as one would naively expect.
- changed
- changed
Comments on v1
Guenther
Indeed, it might be interesting to hear if there are further proposals for kinematic plots? Maybe a procedure could be to produce a set of plots that are given as suppl. material on some Twiki only. Regarding the text (and not repeating the comments sent already by Tristan and AM): l15 : generically instead of generally? changed l58: no comma after "detector" removed title of section 3: I would write : Simulated samples changed l77: I think the MLM matching is relevant for the parton shower, not for hadronization (as written here) you are correct, changed l91: the CTEQ6M PDF set is used at which order? LO, NLO, NNLO? changed to explicitly state cross section calculated using FEWZ at NNLO l212: ... is the simulated fiducial cross section.... changed l248: The uncertainty in the ... changedAnne-Marie
l6 I have been bothered by this a while You spell out LHC and CMS but
not ATLAS changed
l13 note -> paper changed
l22-23 for the paper you need to add the detector section back to the text. added as directed in https://twiki.cern.ch/twiki/bin/viewauth/CMS/Internal/PubDetector
All figures: should now have CMS without Preliminary. changed
Things we can probably discuss in parallel to the LE review:
New fig 4: could rebin the lepton pT distribution above pT>100 GeV ? Or
show in log scale ? agreed
For the choice of distributions to show, I don't know if we want to
limit ourselves or show a representative set for the kinematics of both
jets and lepton, or search variables like Mjj ? Should we have an
ARC-author meeting about this in parallel with the LE review ?
possible distributions are shown above
Tristan
1) What is the motivation to compare the cross-section measurement to these 4F/5F Pythia6/8 predictions of MCFM/MadGraph? The answer is probably because these are the ones that are available... but it would be nice if you could describe some motivation (e.g. around lines 234-247). Currently, the predicted cross-sections in the paper are 0.51/0.51/0.49/0.50, so practically indistinguishable by the measurement... the reader might wonder what we learn from this comparison, and it's also a question one can expect during CWR. The differences in predictions that arise between modelling the b quark as massive or massless are still being understood in general, so we provide insight into this comparison within the context of W+bb. Additionally, the improvements to showering that are brought by Pythia 8 over Pythia 6 could have produced different predictions and having interfaced P6 and P8 with the same MadGraph ME calculation allows us to isolate this difference. In general, different calculations giving similar predictions is of itself an interesting result, and we add our observed cross section to complete the picture. Added:Comparisons between the results of calculations performed
under different assumptions provide important feedback
on the functioning and validity of the techniques employed.
Differences in predictions arising from the modelling of
b quarks as massive or massless are possible, as are
variations in predictions arising from the use of different
showering packages (\PYTHIA6 vs. \PYTHIA8) or matrix element
generators (\MADGRAPH vs. \MCFM).
In the phase space explored here, these predictions are all
very close in their central value and agree with each other
well within their respective uncertainties. 2) You describe around line 81 that "the 4F signal sample is used for the shape.. the normalization from the 5F W+jets sample". It is described quite clearly how the 5F W+bb signal is defined from the W+jets samples, but not so much how it is acutally merged with the 4F sample. E.g.: in my last reading I was wondering if you do you apply the same gen-level selection criteria on the 4F W+bb sample to define this sample. As this merging of two samples is not trivial to do consistently, it might be helpful to give a few more details. The 4 flavor sample is generated as W+bb, and then the same reco-level signal region selections as are imposed for any other sample. To "merge" with the 5F sample, we simply rescale the yield of the 4F (after all selections) to match the yield of the 5F (after all selections). 3) You have now added Fig.4, with some kinematic distributions, which is nice. However, the description is currently still quite minimal (and as AM said: we could discuss which variables to show). The technique for obtaining QCD is slightly different for these variables than for mT, but the procedure is otherwise the same (hence not so much of a description for how we obtained the distributions). For QCD, the difference is that instead of using an inverted isolation sideband for the shape, we use an mT<30 GeV sideband. Then the normalization is set to be that used for the mT variable. As a suggestion, it would e.g. be nice to discuss the data/simulation difference around DeltaR (b,b)~3.5. Note that such a tension was also observed in previous CMS measurements of Z+bb. (E.g. SMP-13-004 and EWK-11-015, but unfortunately I cannot upload attachments in HN now.) Looking at https://cms-physics.web.cern.ch/cms-physics/public/EWK-11-015-pas.pdf
Figure 5 does not show a difference at dR(b,b)~3.5 while some tension is seen in AN2012_303_v7 (I couldn't find dR(bb) distributions in any of the papers) and there the comment is made that this "suggests a harder spectrum in data than expected." In general, the agreement between simulation and data in this analysis is good, and it is hard to draw a conclusion from one data point being 20% high, especially when the points to either side agree at the 10% level. This could be statistics or it could be QCD, but it is not clear what conclusion you would like to be drawn here.
Topic revision: r50 - 2016-10-03 - AlexanderSavin
Home 
Plugins
Sandbox for tests
Support
Alice
Atlas
CMS
LHCb
Public Webs
Sandbox Web 
Create New Topic
Index
Search
Changes
Notifications
RSS Feed
Statistics
Account 
Copyright &© 2008-2020 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