ytatus94 / ewk23l Goto Github PK

Dear Yu-Ting,

Sorry for the delay. These are my old inputs that I used for testing things ages ago, you are probably much better looking at Nicky’s new 3L n-tuples once they are available on lxplus.

However, if you do want to check out the format, a few of the 2L ones are here:
/afs/cern.ch/work/w/williams/public/MultiLepInputMoriond2017

Best wishes,

Sarah

[17.05.17, 10:34:57 AM] Ximo:

// --- Begin extracting and drawing contours
TGraph2D* g = new TGraph2D();
int ret = TH2toTGraph2D(hFinalSig,g);
if(ret==0){
    cout << "no points left, exiting" << endl;
    return;
}

TH2F* Ghisto=(TH2F*)g->GetHistogram();
// Contours for 1.64sigma (~exclusion) and 3sigma 
//   Double_t level[2]={1.64,3};
//   Ghisto->SetContour(2,level);
Double_t level[2]={1.64,3};
Ghisto->SetContour(2,level);
Ghisto->Draw("cont1,list ,same");
// --- END extracting and drawing contours

[17.05.17, 10:35:01 AM] Ximo: with:

int TH2toTGraph2D(TH2F histo, TGraph2D g){

int firstxbin = 1;
int lastxbin = histo->GetXaxis()->GetNbins();
int firstybin = 1;
int lastybin = histo->GetYaxis()->GetNbins();
int npoints = 0;
for (Int_t j = firstybin; j <= lastybin; j++) {
for (Int_t i = firstxbin; i <= lastxbin; i++) {
    if(histo->GetBinContent(i,j)>0 ){
        g->SetPoint(npoints,
                              histo->GetXaxis()->GetBinCenter(i),
                              histo->GetYaxis()->GetBinCenter(j),
                              histo->GetBinContent(i,j));
       npoints++;
    }
}
}
return npoints;  
}

[17.05.17, 10:35:40 AM] Ximo: where hFinalSig is a TH2 with the significance for all your points
[17.05.17, 10:48:17 AM] Ximo: Double_t level[2]={1.64,3};
[17.05.17, 10:48:40 AM] Ximo: I think this is what TH2toTGraph2D() does. -> correct
[17.05.17, 10:49:06 AM] Ximo: Ghisto will have the contours corresponding to 1.64 and 3 sigmas

• For the Higgsino sample, C1 is at the mid-way between N2 and N1

  • If this is the same for the NUHM2, then the Higgsino sample is identical to compressed + compressed.

• We list the NLO cross section: so the k factor should be 1.

• Truth level analysis

  • Generate TRUTH3 derivation

• Check the sensitivity for the compressed + accessible

  • Significant plots

• Request the compressed + compressed official production

• k-factor is one.

• Remove 1 lepton filter

  • But Tommaso suggested to use 2L3 filter (what is it?)

• Use the log file to calculate the 1 lepton filter efficiency

• At low m1/2, 1.2 to 1.6 but it should be 2

• How to generate the mass spectrum?

• ToDo:
  • Produce the TRUTH3 derivation from the event gen level using the following dataset
    mc15_13TeV:mc15_13TeV.370617.MGPy8EG_A14N23LO_NUHM2_m12_300_weak.evgen.EVNT.e4209
mc15_13TeV:mc15_13TeV.370618.MGPy8EG_A14N23LO_NUHM2_m12_350_weak.evgen.EVNT.e4209
mc15_13TeV:mc15_13TeV.370619.MGPy8EG_A14N23LO_NUHM2_m12_400_weak.evgen.EVNT.e4209
mc15_13TeV:mc15_13TeV.370620.MGPy8EG_A14N23LO_NUHM2_m12_500_weak.evgen.EVNT.e4209
mc15_13TeV:mc15_13TeV.370621.MGPy8EG_A14N23LO_NUHM2_m12_600_weak.evgen.EVNT.e4209
mc15_13TeV:mc15_13TeV.370622.MGPy8EG_A14N23LO_NUHM2_m12_700_weak.evgen.EVNT.e4209
mc15_13TeV:mc15_13TeV.370623.MGPy8EG_A14N23LO_NUHM2_m12_800_weak.evgen.EVNT.e4209
    • Can send a request to SUSY derivation or produce by myself
    • TruthDAOD twiki: https://twiki.cern.ch/twiki/bin/viewauth/AtlasProtected/TruthDAOD
  • SimpleAnalysis package: https://gitlab.cern.ch/atlas-phys-susy-wg/SimpleAnalysis
    • This package can produce ntuples, apply cutflows and count the number of events in the SR.
    • Add cxx file in ROOT folder.
  • Calculate the filter efficiency

Dear Judita and Yu-Ting,

Sorry for the delay, its been crazy busy!
Some news I can give you on this though: we are awaiting for some jobs to finish as this I'm afraid is a 2 step process for us if we want to do the normalisation correctly, for this we need to know the number of raw events in the full AOD (we are not biasing with the SUSY2 skimming) per subprocess, so this is where we are at the moment, once that's sorted we can run over the SUSY2 files and do the normalisation on the fly
Sorry it's not as quick as I originally thought but we are getting there as I think the jobs are quick to run on the grid!

Cheers,
Nicky (& Nicola)

• NUHM2 EWK MC had applied at least 1 lepton filter, so we have to check that
  • Does the filter efficiency take into account?
• We can compare the cross sections between Abe's spreadsheet and the SUSYTools.
  • If the cross sections are different between Abe's spreadsheet and SUSYTools, then it implies the filter efficiency had been combined with the cross section in SUSYTools.
  • The filter efficiency can be found from AMI.
• We can ignore the shape fit and RJ (recursive-jigsaw) temporarily.
  • Use cut-and-count to show the significance.

Dear Judita, all,

Apologies for the delay replying- as you can imagine everything is a little hectic at the moment. Anyway, we briefly had a chat with the conveners about the plans for the reinterpretation. Essentially, owing to the tight timescales for the 2/3L paper and the fact it is already a complex paper we cannot promise anything at this point, however we would of course like to encourage you to start studies on this as soon as possible.

As for publication, you may be aware that the recursive jigsaw channels have been descoped from the current paper effort and will proceed to approval ASAP and likely be included in a separate/later paper. This could be an obvious opening in which to include the reinterpretation.

To proceed quickly with this it would be good to see Zn plots (and then HF exclusions) in these signal grids for the signal regions you want to consider. The fits, particularly for 3L, are quite complicated as we do a shape fit so it would be important to show that this works for your grid.

Since we have gone through the full analysis process I imagine it would be quickest to try and get your signal points in the same form as our HF ntuples in order to test the sensitivity. If you don’t have an analysis framework set up I imagine the best thing to do for this would be to speak nicely to one of the groups who have their framework in place :)

How does this sound? If you would like to learn a bit more about how the analysis is currently set up I have some time after the ATLAS plenary meeting has finished, we can of course also talk next week,

Best wishes,

Sarah + Nicky

• Discussion at 10:00 am in R1
• Additional interpretation in EWK 2/3L
  • Use existing SR to show the sensitivity for NUHM2.
  • Make exclusion plots
• EWK group expects to publish their results in LHCP (5/15 to 5/20)
• EPS (July) Higgsino or 0L
  • If no sensitivity for Higgsino, then switch to 0L.
• Judita
  • is working on strong 0L and 2L now.
  • 4/28 to 5/14 not at CERN
  • stays at CERN till 7/28
• Ben:
  • Ask someone to help me to run EWK framework and I run HistFitter using EWK configurations.
• There are several final states (FS) in the weak interaction, the total uncertainties are U_{tot} = \frac{\sqrt{\sum_{i} (U_{i}^{2} \cdot \sigma_{i})}}{\sum_{i} \sigma_{i}}
• Pat:
  • Bear's paper says SS di-leptons EWK not 3L.
  • But the 8 TeV results are 3L. (2L is not exclusive).
  • We don't study multi b-jets channel because the sensitivity is small and the behaviour similar to SS.

• Sarah:
  • EWK2/3L approval meeting on next Tuesday.
  • Use comment configure file for the HistFitter ntuples.
  • Read the appendix of the supporting note.
  • EWK 2/3L group meeting: every Thursday 3:30pm
  • 2 approach
  • SUSY2 p2879 is fine.
• Higgsino extends to lower and higher $\mu$
• New grid points in compress region.
• HistFitter ntuples
  • From Sarah
  • Bkg and signal
• Significant plot
  • Give a talk in EWK 2/3L group meeting when significant plots are ready
  • Maybe need to add N-1 distributions
  • Set HistFitter to run exclusion plots
  • Give a talk again.

1. Prepare the sensitivity plots
   • Treat the objects in ntuple as signal objects apply the necessary cuts on the objects.
   • The cuts also can be found from the HF configurations
   • Use signal and background to calculate the significance (Zn)
   • Use all background samples (the number can be obtained from the note)
2. Run HistFitter
   • Run 2L configurations after we get the common ntuples
   • Prepare the macro for the exclusion plots and make the plots
   • Run 3L configurations
3. Judita will be away next two weeks.
4. Is it pure Higgsino?
5. We didn't generate curly compress state.
   • with highest cross section
   • single jet + 2/3L
6. Truth level study

Dear Yu-Ting, all,

Since you are, as I understand, more interested to begin with at looking at the 3L channel, I am cc-ing Nicola and Nicky in who have been working on the fits in the 3L channel. As we discussed a few weeks ago, most groups in the 2/3L analysis have their own framework (and so extensive outflow comparisons have already been undertaken) but for limit setting we have a common ntuple format, which you can see here:

https://svnweb.cern.ch/trac/atlasphys-susy/browser/Physics/SUSY/Analyses/WeakProduction/HistFitterTree/trunk

It would be, I think, relatively easy to write a piece of code to extract yields based on these ntuples and hence produce the sort of optimisation plots you would need to demonstrate sensitivity.

For HistFitter we also have a common config in the limit setting package:
https://svnweb.cern.ch/trac/atlasphys-susy/browser/Physics/SUSY/Analyses/WeakProduction/Limits/trunk/13TeV/2L
(Look under ConfigFiles/*Moriond.py,
though if I understand correctly the 3L part is not completely up to date on svn.

My feeling is, in the interest of efficiency, if Nicky/Nicola had time to run these signal points through their framework and make the HFT ntuples, then you could try running of those. I know both are very busy with the current paper effort though so alternatively I could probably as a starting point try to run something but it would be with an older release.

In the meantime in order to get a feel for sensitivity, provided you have any analysis framework that could roughly reproduce the selections needed then you could also start just looking at the Zn values, since all of the background estimates are in the supporting note,

Best wishes,

Sarah

• Use the bkg sample to prepare the code for cutflow comparison.
• Make the significance plot as what Ximo did.
  • x axis: m12, y axis: nothing, show the Zn values on each grid point.
  • Use all the signal regions defined in the EWK 2/3L supporting note.

Hi all,

Thanks for starting to look into this!
Regarding the configuration of the 3L part, I plan on uploading this at some point in the next few days but in the meantime we can certainly run over the signal points
On this note, is there any xsec info for them?

Cheers,
Nicky (& Nicola)

• 20170424
  I request SUSY16 with p2952 for NUHM2. Here is the JIRA ticket:
  https://its.cern.ch/jira/browse/ATLSUSYDPD-879?filter=-2
  I confirmed with Joey that p2952 is the current tag they are using now.

Hi Sarah,

Sorry for replying just now, I was at a meeting. Thank you so much for providing us the material! Therefore we will use:

1. Background samples taken from the HistFitter config file as
   bgdFiles.append("/usera/sarahw/testareas/PostPHD/TwoLepHFProcessingMoriond2017/HFT_Central_110217.root”)
   dataFiles.append("/usera/sarahw/testareas/PostPHD/TwoLepHFProcessingMoriond2017/Data_CENTRAL_no2L2J.root")
   fakeFiles.append("/r03/atlas/sarahw/HFInputsMoriond2017/central_Egamma_2L.root")
   e.t.c
2. HistFitter configuration file (indeed I am not finding the 3L part, but having the 2L configuration is very good for now)

This will be a good starting point, and Yu-Ting can already start with his preparations and use the HistFitter. Next would be to:

1. Have ntuples for the HistFitter of NUHM2 model, and it would be great if Nicky or Nicola would have a bit of time to run them (it is just 10 points).
2. With this Yu-Ting can do the sensitivity study, and provide significance plots.
3. HistFitter update for the 3L analysis, and doing the exclusion plots with CLs. I believe this will be just a matter of uploading it to svn, so I would ask Nicky or Nicola for this. Then Yu-Ting can run them and prepare the nice paper quality plots for exclusion.

Many thanks to everybody, I hope we are not distracting you too much.
Sincerely,
Judita

• 20170421

[21.04.17, 5:33:29 PM] Yu-Ting: Hi Joey, can I disturb you? Do you work on the EWK 2/3L group? Because I saw Penn is involved in this group and I also saw your name.

[21.04.17, 5:35:25 PM] Joey Reichert: Hi Yu-Ting, I do. I might be a bit slow to reply, but feel free to ask away!

[21.04.17, 5:36:40 PM] Yu-Ting: Thank you. Then I have some questions related to EWK 2/3L group's study. Because the SS/3L analysis is almost done, and I discussed with my advisor about what's the next, we decided to work with EWK 2/3L group. Just add our non-universal higgs model (NUHM2) as an additional interpretation. We had discussed with Ben, Sarah, Nicky about this. Finally, they agree to let us add our NUHM2 as an additional interpretation. Since the EWK 2/3L studies are closed to finish, so our schedule is very very tight. But I still not clear about the procedure of the EWK 2/3L study. Sarah told me that each sub-group has their own framework and you used your own framework to produce the common ntuple for the HistFitter. Nicky agrees to help us to produce the HistFitter ntuple of our signal points. I checked the twiki and I saw the spreadsheet for the cutflow comparison. My first question is: Do you use SUSY2 DAOD files as input of your own framework, applying the cutflow, and produce the HistFitter ntuple? If I get the HistFitter ntuples produced by Nicky, then do I need to apply the cutflow on it?

[21.04.17, 5:54:16 PM] Joey Reichert: yes, we do use SUSY2 as input to our own framework. but right now, we (penn) aren’t doing much of the HistFitter ntuples. But you will need to still apply the standard cleaning cuts on the HF ntuples, i believe

[21.04.17, 5:56:39 PM] Yu-Ting: Do you mean penn has different ntuple format?

[21.04.17, 5:57:05 PM] Joey Reichert: for the most part, yes. but this is largely because only nicky and sarah are doing the final HF stuff

[21.04.17, 5:59:42 PM] Yu-Ting: If I get the HF ntuples from Nicky, then I need to apply the standard cleaning cuts. After that, I can use the events passing the standard cleaning cuts to calculate the significance. Is it correct?

[21.04.17, 6:00:53 PM] Joey Reichert: yes, i believe so anyway!

[21.04.17, 6:01:08 PM] Yu-Ting: And after showing the sensitivity, I just need to use the same HF ntuple to run HistFitter to make the exclusion plots.

[21.04.17, 6:01:49 PM] Joey Reichert: i think that makes sense to me

[21.04.17, 6:03:53 PM] Yu-Ting: Ok, thank you. In SS/3L group, Ximo produces the flat ntuples for the group and we use the flat ntuples to do the cutflow comparison, calculate the significance, define the SRs. But the University of Frieburg has their own ntuple format, and they help all the group to do the HF studies. So I was confused about the EWK 2/3L group's procedures. Thank you for explaining things to me. So I think I cannot do anything before I get the HF ntuples from Nicky. I may have more questions about the EWK 2/3L studies in the coming weeks. Could you please spend some time to answer my questions? Thank you in advance. :)

[21.04.17, 7:55:46 PM] Joey Reichert: No problem! And sorry for my slow responses, I actually got into an impromptu meeting right after you started asking things! Probably Nicky and Sarah will have better answers to any issues you have than I will, but I can try to help still!

Dear Nicky and Nikola,

I was wondering if you had a bit of time to produce the NUHM2 samples. Yu-Ting is already preparing his code for the format used in the 2L analysis. There we already have the background samples, but what is still needed for the first significance plots, and later on the HistFitter setup, are the NUHM2 input files.

As there is very little time to finish this interpretation, it would be nice if you could run them. For simplicity, I am listing them here again:

mc15_13TeV.370617.MGPy8EG_A14N23LO_NUHM2_m12_300_weak.merge.DAOD_SUSY2.e4209_a766_a821_r7676_p2879
mc15_13TeV.370618.MGPy8EG_A14N23LO_NUHM2_m12_350_weak.merge.DAOD_SUSY2.e4209_a766_a821_r7676_p2879
mc15_13TeV.370619.MGPy8EG_A14N23LO_NUHM2_m12_400_weak.merge.DAOD_SUSY2.e4209_a766_a821_r7676_p2879
mc15_13TeV.370619.MGPy8EG_A14N23LO_NUHM2_m12_400_weak.merge.DAOD_SUSY2.e4209_s2608_r7772_r7676_p2879
mc15_13TeV.370620.MGPy8EG_A14N23LO_NUHM2_m12_500_weak.merge.DAOD_SUSY2.e4209_a766_a821_r7676_p2879
mc15_13TeV.370621.MGPy8EG_A14N23LO_NUHM2_m12_600_weak.merge.DAOD_SUSY2.e4209_a766_a821_r7676_p2879
mc15_13TeV.370622.MGPy8EG_A14N23LO_NUHM2_m12_700_weak.merge.DAOD_SUSY2.e4209_a766_a821_r7676_p2879
mc15_13TeV.370623.MGPy8EG_A14N23LO_NUHM2_m12_800_weak.merge.DAOD_SUSY2.e4209_a766_a821_r7676_p2879

Many thanks in advance,
Judita

• Use the background ntuple to make the framework for cutflow
• Make the significance plots
  • Use SRs from EWK 2/3L to see which SR has the highest the significances.
• Read section 2 and 5 of the HistFitter note.
• Ben said EWK group might need help.
• Remind Pat to chat with Judita before 5:00pm to discussion about the abstract and Higgsino for next Monday Higgsino meeting.