DVCS with Recoil Analysis Page DUMP

Aus HERMESwiki
Zur Navigation springen Zur Suche springen

Page maintainer: Caro

Workingpage.jpg This is a working page which will continuously change. The content is not reviewed and is intended for people working on this specific topic.

About this page

This page contains all old (processed) content of the working page DVCS with Recoil Analysis Page. By doing so, the parent page is kept clearly structured.

After second circulation

Note: not all processed points have been crossed out.

General

* Prepare answer-collaboration file (Caro)

* Formulate the result of the pure sample in a more positive way. Do not put the reference sample on the same footing as the pure. Put more emphasis on the pure sample.

    • Mission accomplished?

  • Think about exchanging the order of Fig (pure + theory) and Fig (pure vs. unresolved-reference vs. unresolved).
    • Is this easily feasible from the logical flow of the text?
    • Would also help in putting less emphasis on the unresolved samples
    • Introduce an extra section for the triple panel? Would have very little text. Could more content be shuffled there? Call sections e.g. "Comparison with previous measurement" and "Comparison with theory"

  • Improved quantified statement about pure vs. unresolved ref. (this is our physics message). Compare fully correlated sample by applying Mx2 cut on top of pure sample. Sergey.

Text

  • Use JHEP style. JHEP style OK?
  • [CKR 17.10.2011] Why do we not mention the lower thetagammagamma cut and the upper Epre cut? Added.
  • [6.1.2012] Angular resolution is not discussed at all in paper. Add text offline.
  • Cite previous work in the introduction, but only GPD H at line 48.
  • References: get tex file from DC90
  • References: cite PhD thesis for Recoil detector => only if recoil technical paper is not yet published.
  • References: is there a recent review article citing all DVCS measurements on the market?

Figures and Tables

[last update: 15.02.2012]

  • Add a figure DVCS/BH Feynman diagrams (see e.g. DC90)
  • Add a figure diagram with phi-angle definition. Add Trento convention to running text.
  • Add a figure MX2-distribution with histograms for pure, unresolved-reference, and unresolved samples. No MX2-cut on either of these. Show also MC contributions?
  • Add a figure t_c minus t_RD for the pure sample
  • Modify Fig. 1 (Recoil scheme). Make consistent with figure caption; change labeling, add e.g. PD, change set of track configurations shown.

  • Combine Figs 6 and 7 into one: show three data sets in one figure. This will put less emphasis on the unresolved samples.
    • Jen's new plot as of December 20, 2011
    • 3 error bands above each other (no inner/outer error bars)
    • pure at its avg kine; unresolved reference at the avg kine of the pure, but shifted to the left; unresolved at the avg kine of the pure, but shifted to the right.
    • Common caption
    • Reference to table with avg kine in caption
    • Modify running text accordingly

  • Should we use a different / additional representation for Fig 6 (pure vs. unresolved reference)? Showing the difference of the asys, instead of them separately?

  • Plot captions
    • Add shifted for better visibility where needed

  • Plots cosmetics
    • GENERAL: Make figures in the same style (HERMES)
    • xlabels to the right (not centered)
    • vertical alignment in Fig 4
    • do not use different fonts
    • ... Sergey? you made a more complete list on Dec 9, right?
  • Check consistency of Tab 3 and Tab 4. Values for A_final (= corrected result) should agree since Bg correction is the only correction applied.
  • Table 3 (background)
    • Would be very useful to see the extracted A_SIDIS asymmetry in each bin. We concluded that it would be useful to add the actual values of (the integrated) A_SIDIS into table 3.
    • What is the delta excl stat column about in table 3? Is it supposed to indicate an uncertainty or the asymmetry?
  • Need table of all systematic contributions in every bin.

Data Analysis

[last update: 04.05.2012]

  • Photon energy used
    • Mail sent to dvcs-list on April 30, 2012
    • DC92 proposes to use the uncorrected photon-energy values from the smCluster table for the analysis of the pure, the unresolved-reference and the unresolved samples, instead of applying the correction formula that depends on the energy deposition in the preshower detector. We will remove the currently used upper cut on the preshower energy (120 MeV), while keeping the lower cuts on the energy depositions in the calorimeter (5 GeV) and the preshower (1 MeV), respectively. By doing so, we will be consistent with ALL previous DVCS publications, including the latest, DC90.
    • We have discussed this intensively in the core analysis team and in the drafting committee. On the one hand, using the uncorrected values means in a certain sense a draw back in the time error pointing to the future. A lot of effort was in the past put into extracting the two-dimensional dependence of (E_calo, E_preshower) and into determining a parameterization that allowed to correct for the observed slope in the 2D distribution.
    • On the other hand, there are several arguments that speak in favor of using the uncorrected values, which we will summarize now. Ideally, we should have discussed this earlier, but better later than never.
    • The need to change the analysis strategy emerges from the demand to have a decent data - MC comparison. In the current situation (using the correction formula and applying a 0.97 factor in the MC, or smear the generated value), the width of the Mx2 distribution is narrower in data than in MC, and the peaks do not coincide. In the data - MC comparison of the TTSA paper (2008) was much better (apart from the known 10% overshoot of the MC over the data, and the fact that the positron beam data is slightly shifted wrt the MC). The collaboration desires to see a data - MC comparison in the publication, showing all three exclusive samples, and we agree that this is necessary since, last but not least, the purity and selection efficiency of the pure sample depend on the quality of the comparison. We have recently obtained results without using the correction formula on data, and the agreement with MC is good enough.
    • As a side remark, recent event re-weighting attempts to correct the MC in order to make it match with the data failed. Most important, these attempts lacked a detector-based motivation.
    • The 2D (E_calo, E_preshower) distribution in MC is known to be different in data and MC. In an attempt to correct the MC in a similar fashion than the data, a correction formula was extracted for MC in the past, yielding different pre-factors, but with the same functional dependence on E_preshower and E_calo. This correction formula was tested and was shown to severely overcorrect the MC, by making the peak width 40-50% narrower than in data (with data correction formula). This probably points to the fact that the correction procedure is not the optimal one. It also relies on the applicability of using the correction derived for leptons for photons, which might be momentum-dependent. Re-iterating the correction formula is out of reach, for reasons of manpower and the strong wish to close certain lids. The current correction formula is the best knowledge we have, but there is strong indication that this knowledge is not good enough.
    • For the current publication, we could probably find arguments to justify the bad data-MC agreement since we showed that the purity (extracted from MC) is only very weakly sensitive to the photon-energy calibration used. Still, in view of the upcoming "associated paper draft", we have to think further. For this paper, it will be much more important to have a really good simulation at hand since the fractions of involved background processes are larger and are as such essential for the physics message of the paper. Certainly, we should use the same analysis policy in draft 92 and in the associated paper.
    • Having said all this, we conclude that we prefer to use an unbiased photon energy for the kinematic event fit, rather than a correction of which we do not know how much we can trust it.
    • Lastly, the advantage will be that this approach will be consistent with all previous HERMES DVCS publications. This argument is not to be neglected since we spent quite some effort in making sure that the new results are backward compatible.

Remark: the cut on the chi2 from the kinematic event fit will be re-adjusted (in any case checked). The distribution of the probability should be flat, ideally.

  • New data productions (06f and 07d).
    • Check whether time dependence of DVCS yield and missing mass is absent when using single-cluster algorithm. If yes, missing-mass shifts can be entirely removed. (we have only one beam charge, no need for shift between positrons and electrons)
    • Re-extract asymmetry amplitudes
  • ML fit of asymmetry
    • Check also sin3phi.
    • Check correlations between sinphi and sin2phi. See DC90 and DC91 correlation plots.
  • MC-data comparison (30.1. / 10.2.2012)
    • Need to be careful with the percentages quoted (purities, efficiency of pure events selected). Depends on MC generator, associated fractions, detector inefficiencies. (how much?)
    • Check how good data-MC comparison was earlier: checked, TTSA plot looks very similar. How good is the ratio in our April-2011 release report?
    • Plot data-MC comparison for x,Q2,t. Jen.
    • Plot purity in individual bins of x,Q2,t. See release report!
  • Associated fraction
    • Try to estimate associated fraction in the recoil acceptance from real data (see e.g. Sergey's presentation Dec. 2011). Study to be continued.
    • (very maybe) Alternative asso fractions from alternative DISng - can this be used as sys uncertainty on asso fractions? (see documents from a study in summer 2008 compiled here) What sense would a systematic uncertainty make?
  • Study traditional exclusive cuts
    • -t
      • Check -t distribution of pure sample (only). Remove upper -t cut?
      • Goal: add one more bin? Check statistics in bin 0.7-1.0 or 0.7-2.0 to the stat in the other t-bins.
      • Could be added as open point in only the t-dependence, without including the data t>0.7 in the x- and Q2-dependences.
      • [Jen summer 2011 - asking M. Diehl] Extend the t cut? My understanding is - fine, but need to be careful when comparing against GPD models - not so good at lower Q2, but we could for the higher t have a higher low Q2 cut of about 2 if we want to compare the higher t to gpd models, or have a cut off point in t if we compare to gpd models ]
    • ThetaGammaGamma
      • Remove upper thetagammagamma cut in pure sample. How much statistics is gained? Keep lower cut (is resolution-related).
      • Is the upper thetagammagamma cut needed to have flat phi-distribution? Check Dietmar's PhD.
      • MC studies of purity and efficiency should be repeated. Associated contamination? Basically whole analysis has to be repeated.
      • Goal: increase statistics
    • Originally we kept the traditional cuts in oder to keep compatibility with the traditional, unresolved samples
  • For us (not to be included in the paper): compare generated amplitudes pure vs. unresolved, and reconstructed pure vs. unresolved.
  • Detailed study of acceptance functions. Study p-cutoff effect on ref. sample. (goal: systematics due to acceptance function). Related: compare kinematic distributions between reference & traditional.
  • Inefficiencies in Recoil
    • False asymmetries arising from time-dependent inefficiencies in the Recoil.
      • Charlotte checked the inefficiency studies Xianguo performed.
    • Another possibility (apart from assigning a systematic uncertainty) could be to correct for the inefficiencies (XTC level or by weight in the event list for the ML macro).
    • Combine inefficiencies and reference sample in such a way that consistency can checked.
    • (Abandoned) Study ineffs of SSD. Still needed? Eff. for protons is close to 100%. Dead strips are modeled in MC. Exclude dead strips in MC and compare?
  • Detailed misalignment studies
    • involving all 3 detector systems (front tracking system, calorimeter, recoil). Misaligned MC production?
    • WD's mail of April 7, 2011. Caro makes t-test from plots shown at release talk (?)
    • Can misalignment affect purity? chi2 distribution can be shifted differently in the pure and the associated event samples. However, little effect is expected on the purity. Rather the efficiency is expected to be affected.
  • Measured vs. fitted parameters. Which kinematics will we use in the end (x,Q2,t, also thetagammagamma, Mx2)? Measured ones from HTC or fitted ones?
    • 15.12.2011: Plot t_constrained - t_after_fit (Sergey)
    • For t, we have 4 choices: spectrometer t (not considred), t constrained, t calculated from Recoil detector: measured or fitted
    • 23.6.2011: argument to use measured parameters: comparison with traditional analysis. Plus we cannot be sure to not catch unestimatable systematics when using the fitted values.

Systematic uncertainties

[last update: 15.12.2011]

  • Aram can maybe help with the systematics.
  • We need a file from Jen with all systematics compiled, for each bin, and split up into the individual contributions
    • Would be very useful to see the extracted A_SIDIS asymmetry in each bin.
  • In general, we plan to re-do the 3-in-1 for all three samples once the gmcDVCS MC production with higher statistics is available. (??)
  • Pure sample
    • 3-in-1: is OK modulo adding more stat in the future
    • SIDIS and excl pion accounted for, but very small and negligible. Is OK, no need to be re-done.
  • Unresolved reference sample
    • Jen's analysis not 100% x-checked
    • 3-in-1: for the April-2011 release, Caro/Sergey gave reconstructed MC asymmetries, Jen ran qplot and calculated the difference = sys. Must be re-done.
    • SIDIS corr: NOT done. Plan: take from unresolved sample.
    • SIDIS sys: calculated by Jonathan for the April-2011 release, but the rec 2-photon asy was not sent through the acceptance function at that time. Plan: take from unresolved sample.
    • NO Mx2 shift was applied for the April-2011 release (will very likely become obsolete for the final publication if using new productions). Therefore (in v1.0) the sys error band is smaller compared to the unresolved sample.
  • Unresolved sample
    • 3-in-1: done by Jonathan for the April-2011 release using ALL 2006/2007 runs. Must be re-done.
    • SIDIS corr: NOT done
    • SIDIS sys: calculated by Jonathan for the April-2011 release
    • Mx2 shift was applied for the April-2011 release (will very likely become obsolete for the final publication if using new productions)
  • Model used for 3-in-1
    • VGG / gmcDVCS is used now, using all 5 variants and building the average
    • Get also result for gmcDual, add to the 5 VGG results and average the 6. Check whether it gives difference. (??)
    • Pro VGG-only:
      • gmcDVCS describes the pure sample pretty well. Then we should also use gmcDVCS for the other two samples, in order to be consistent
      • gmcDual does not describe the HERA collider data
      • gmcDual has only one variant, gmcDVCS has 5.
      • If including Dual in the systematics does not result in a significant difference, don't use gmcDual. But: I (Caro) am not sure whether we have the time & manpower to realy go into this exercise.
    • DC92: VGG matches better the main physics result (from pure sample) than does Dual.
    • DC90: Dual matches better the main physics result (from unresolved sample) than does VGG.
    • What DC90 uses for systematic uncertainties (3-in-1): gmcDual with the old geometry and the not corrected version (factor of 2).

Theory comparisons

[last update: 15.12.2011]

  • VGG: use only one model variant, the one that is closest to the measured data points, instead of showing the (not so nice, and potentially in the overall bin misleading) fat band
  • Müller / Kumericki http://calculon.phy.hr/gpd/ Jen: have you meanwhile heard back from them about the lowest x- resp Q2-bin?)
  • Liuti : Jen will contact her ??
  • What DC90 uses for theory comparison: Müller / Kumericki and Liuti. VGG is out
  • [Sergey 22.8.2011] Instead of comparing our measured data to the model in 4pi, can't we compare (in the plot) our data to the model in our acceptance? Related to this: our all-in-1 uncertainty related to acceptance is actually an asymmetric error; however, we include it in our sys band as symmetric one.


After first circulation

Our Notes 2011/11/25 -more ordered and cleaned on the main page

  • General
    • ! If we have the chance to formulate the result of the pure sample in a more positive way, we should take the opportunity.
    • ! Quite some general comments - do not put the reference sample on the same footing as the pure. Put more emphasize pure sample.
    • ? why are error bands in Fig. 7 so different in size? Answer by Jen: In figure 7, no systematics for the missing mass shift is included for the reference sample, which gives us the big difference. => with new prod, rerun over data with multi-cluster algo. Check whether missing-mass shift is needed at all.
    • Estimate of sys should be done with higher statistics gmcDVCS. Dietmar used much more stat.
    • Should include sys in comparison? If yes, no need anymore for reference sample - results would not differ. see answer to Marco ==> One cannot use the systematics in the comparison of the 2 samples. They are highly correlated and they account for acc effects that go well beyond recoil detector
    • In general, the DC is against using Dual for systematics - we would like to stick to what we have now.
  • Style & add info (easy):
    • Take template of JHEP from wiki webpage. Check JHEP style.
    • Make figures in the same style (HERMES)
    • Cite PhD thesis for Recoil detector => only if recoil technical paper is not yet published.
    • Cite previous work in the introduction, but only GPD H at line 48
    • Add in Table 1: <x>, <Q2> for each <t>-bin etc.
    • typo in first xB bin for pure
    • Add figure of missing mass with all three samples
    • Ask Jennifer to repeat plots without any shifts "for better visibility"
    • We have to add to the panels, in case we keep the shifted version, "shifted for better visibility"
    • plot t_constrained - t_after_fit (Sergey)
    • Check consistency of Tab 3 and Tab 4. Values should agree, right? Bg correction is the only correction applied
  • Major to semi-major analysis items:
    • For us: Mc_test_1: compare generated amplitudes pure vs. unresolved, and reconstructed pure vs. unresolved
    • Sergey will look at it (maybe with help from Jen for qplot)
    • Check -t distribution of pure sample (only). Remove upper -t cut? Add one more bin?
    • Could be added as open point in only the t-dependence, without including the data t>0.7 in the x- and Q2-dependences.
    • Remove upper thetagammagamma cut in pure sample. Keep lower cut (is resolution-related).
    • MC studies of purity and efficiency should be repeated. Associated contamination?
    • Basically whole kinematics has to be repeated.
    • Goal of thetagaga: increase statistics
    • Originally we kept the cuts in oder to keep compatibility
    • Try to estimate associated fraction in the recoil acceptance from real data [Sergey]. What if we see different results?
    • (very maybe) Alternative asso fractions from alternative DISng - can this be used as sys uncertainty on asso fractions?

Paper first circulation

* Add more text about the recoil detector in the experimental section. Sergey

    • Track finding, fit, rdPID.

* Some footnotes / remarks added in italics or red in the text need to be fixed. Caro

    • e.g. text has to be added about fit results without constant term in section 5.
  • Add / replace figure showing correlations Recoil - Spectrometer. DeltaPt and DeltaPhi, each one-dimensional. Real data pure event sample. Sergey
  • Fill tables of results and average kinematics. Jennifer

* Figure style: in progress. Jennifer

Figs 5 & 6: need estimate of sigma deviation for 1. pure vs. unresolved ref. & 2. unresolved ref. vs.unresolved. Use diff and sqrt(delta1^2 - delta2^2), i.e. assume fully correlated data sets even this is not 100% true; will overestimate possible effect --> "conservative estimate". Compare with Sergey's results of March/April, where he applied the Mx2 cut on top of the pure sample to generate truly 100% correlated samples. Caro

  • (+) The background section might the appropriate place to add some literature hints (...) about the associated asymmetry. next publication
  • (+) Find out whether for us regarding resonant production, only the Delta+ is relevant? Maru text is written carefully enough now.
    • (+) Adding lines of summary of Sept-19-meeting: Remarkably, it was only HERMES throughout all those years that made a strong point out of the Delta contamination. We therefore have to be careful when talking about DVCS measurements by other experiments. For me (Caro) it is at this moment not clear whether at JLab, the Delta contribution is simply lower, or whether their resolution is simply better (even when not detecting the photon), or whether it is a convolution of the two. It might a point we should find out. (maybe we are just in an unlucky regime of the Delta x-section) Answer later 2011: they have better resolution and can separate the associated off even without detection of all 3 particles
  • (+) Sample fractions (table) Caro Dec 2011: the fraction table died meanwhile...
    • Comparison real-MC data does not necessarily have to be perfect - if real data is not helicity balanced, numbers may be different. Also, differences in fractions between helicity states can resemble physics - asymmetries for a, b, c are different! Add uncertainties of fractions (they are very small though...).
    • real data: weight 1 term such (with fraction) that sample gets beam-heli balanced.
    • add numbers for ineff. MC (gmcDVCS)
    • add somehow DISng (mix in appropriately - have to think how I will technically do this)


    • DVCS yield vs. time for 06e1 and 07d1 productions, single-cluster algorithm (Caro Sept. 2011)
    • DVCS yield vs. time, comparing 07c2 multi-cluster (release April 2011) with 07d1 single-cluster algorithm (Caro Sept. 2011)

DC92 Meetings before April-2011 release

Skype Meeting 17.3.2011

Next Meeting: Sat March 19, 13:00 HH time. Deadline for version to be discussed is 9:00 HH time on Saturday. Version 0.11 to discussed
Agenda: Minutes in italic.

  • General remarks?
  • Reference sample: resolved. To be done: Fractions elastic, associated, sidis (Caro) and sys. uncertainty all-in-one and sidis (Jen/Jonathan).
  • Time dependence: progress? Compare time bins (1-4) and (2-3). WD's idea: is it the beam energy that is the culprit?
    • Jen looks into it on Friday
  • Plots for release: status?
    • Caro and Sergey will send reference amplitudes.
  • Status of systematics: no sys for HTC/HRC; LD/HD; quadrants; H0; RD ineffs; ?sidis in elastic?; ?upper chi2 cut?. We do have: all-in-one; and we should have: time-dep. No sys at all due to Recoil?? They will not buy this.
    • We have two classes of sys studies. Make this clear in release talk.
  • 1. FIN
    • Morgan will today modify a phrase on comparing our results to publications.
  • 2.1 FIN
  • 2.2 (Recoil track info) TUNE
    • Sergey will finish it today
  • 2.3 (The elastic sample) TUNE -> FIN
    • Caro will have finished fractions by Friday.
  • 2.4.1 (Mult, ghosts and clones) FILL
    • Do not include the asy plot, keep in backup.
    • Sergey looks into it. If there is a result by Saturday, put it into report, if not, study in remaining 2 weeks.
  • 2.4.2 (KinFit Opt) PREP
    • Main topic of this section would be: fitted minus measured parameters.
    • Don't change approach that measured kine are used for release.
    • Sergey observes difference if using measured or fitted in exp. data, but not in MC.
    • DC agrees to remove this section in the release report and study in greater detail in the near future.
  • 2.4.3 (Upper chi2 cut) TUNE -> FIN
  • 2.4.4 (rdPID) PREP
    • Status should go to FILL or even TUNE
    • MC mentioning should go
    • Caro will send file
    • Some of the red text will be integrated in 2.2 (Caro)
  • 2.5 (Reference) TUNE/FILL (more on C_A? What does reference answer look like for associated events?)
    • Sergey works on associated.
    • For the reference it is our first goal to produce similar average kine for ref and elastic.
    • add that half is going by phi, half by p rdacc Fig.2.17
    • right plot of fig.2.18 goes away + "complementary situation
    • instead, add C_A plot by Sergey in fig.2.18
  • 3. (Asymmetries) FILL (more text on add. harmonics in text.)
    • Jen will tune text on add. harmonics.
    • People are in favor of not including asso plot. Prepare for collab meeting though. Put phrase into slide what is the plan for bg asymmetry (draft by mail)
    • Change status to TUNE
  • 4.1 (all-in-one) TUNE (please tune this text)
    • Morgan will work on chapter 4 this afternoon
  • 4.2 (Time dep) TUNE
    • put to FILL
  • 4.3 (LD/HD) TUNE
    • Morgan tunes
  • 4.4 (HRC/HTC) TUNE
    • Morgan tunes
  • 4.5 (H0) FIN
  • 4.6 (quadrants) FIN
  • 4.7 (Rdineffs) TUNE
    • Morgan will read it
  • 4.8 (upper chi2-cut) PREP (maybe goes away completely. Caro will discuss with Ami)
    • will move to later. Remove.
  • 4.9 (Background) TUNE (please merge with App.A)
    • App A merging done, App B removed.
    • Caro will send event list including beam pol
  • 4.10 (Summary of sys) FILL (fill!)
  • 5. (Results) FILL
    • Caro will fill text
  • 6. (Summary) FILL
    • Status is TUNE
    • needs tuning after Sat. meeting

Skype Meeting 14.3.2011

Version 0.8 to discuss
Agenda: Minutes in italic.

  • General comments by anyone?
    • Our internal deadline is Friday, March 18. This means we have to FILL and TUNE !!
    • Interactive writing was very successful last week. Once section finished, could be read by WD.
    • Major identified open points: discussion on ghost tracks; reference sample; systematics due to time dependence. (see below)
  • Start filling tables of systematics! (added at the end of systematic chapter).
    • Need to give info on individual contributions vs. overall, t, Q2, x
  • Time dependence. Jen's plot of March 11. Suggestion: study other/more time bins.
    • Need to understand origin of time-bin 2 constant term. Most of us agree that the 3 sinphi time bins agree within stat (are independent data sets)
    • Combine 1-4 and 2-3, i.e. periods with similar yield. Difference can be assigned as sys.
  • Reference sample: Sergey (Andreas' acceptance function) and Caro (Dietmar's acceptance function) are converging. We are currently looking into the details of our differences. The text of the reference chapter is written and ready for inspection. We are aware that Glasgow needs the event list of the reference sample to calculate the traditional systematics. This is possibly something that can be done also during the circulation of the report.
  • Status of inefficiency studies (Sergey) included in report.
    • Is good estimate already.
  • Sys due to SIDIS in pure elastic sample
    • Status: extraction has started.
  • Sys due to upper chi2 cut
    • Caro will look into Ami's thesis, then will get in contact with Ami.
  • Results chapter: Caro offers to tune text (which is still blue) in the next few days
  • Extra meeting next Friday, March 18?

Skype Meeting 7.3.2011

Agenda: Minutes in italic. The present DC members agreed on a regular skype meeting on Monday, 13:00 HH time for the time being. Next meeting: Mar 14

  • General remarks by anyone?
    • Text writing needs more progress until next week!! Produce content and send around a mail as soon as a section reached the status "TUNE"
    • Start preparing and filling table of systematics!
  • Progress on writing
    • WD worked on introduction. Offers to read certain chapters. Glasgow will write systematics and other chapters foreseen and provide more content until next Monday. Caro will concentrate on writing her sections
  • Status of cross-check
    • Pure elastic x-check considered to be done by DC. Displace points more.
    • Caro needs some more numbers and amplitudes from Sergey and Jen to put into the report.
  • Status of reference sample (Caro, Sergey).
    • Sergey looked into alternative rdacc() function. Analysis summary and an attempt of final conclusion by Caro by the end of the week.
  • Status of rdPID studies, Jen's latest plot
    • Asymmetries are very stable under rdPID cut. This is an indication that the rdPID cut is basically redundant after the event reconstruction by kinematic fitting.
  • Status of inefficiency studies (Sergey) /user/yaschenk/Release1104/gr_asym_tc_*.eps Sergey's plot
    • Are we sure the difference in central value is independent of the MC stat? Can more MC be produced easily? Is the difference = sys considered to be negligible?
  • Systematics due to time dependence (everyone). Check asymmetries in different time periods. Progress?
    • Glasgow will look into time periods probably this week. How to choose time periods? Idea: 4 equally large time bins in 2007, combining quarter 1 (-) and 3 (+), and 2 (-) and 4 (+). Extract 2 independent pure elastic beam-helicity asymmetries (+)-(-)
  • Systematics due to chi2 cut -> contacted Ami
  • Status of traditional systematic uncertainties (Glasgow). Reminder: Glasgow needs an eventlist for the reference sample from Hamburg.
    • All-in-one not yet done with gmcDual - MC production pending. However, elastic amplitude from real data does neither resemble wrong gmcDual nor the corrected gmcDual (lies in the middle). The DC agrees to use gmcDVCS for the estimation of the "all-in-one" uncertainty as best available toy model. A plot comparing data and MC is the latest version of the report.
  • Status of HTC/HRC comparison plots (Jen)
    • Plots available, will be put in report
  • Plots for release (Jen): on the meeting on Monday, 28.2., the general consensus was to prepare for release a) pure elastic b) pure elastic + reference c) pure elastic + reference + traditional all 2006/2007 (DC90 except for no separation of interference and squared DVCS term, which means that someone (Jen) needs to run Jonathan's event list through the EML macro for cross-check reasons). An alternative to c) would be cc) pure elastic + reference + traditional 2006/2007 subsample with fully operational Recoil detector.
  • Major identified open points: discussion on ghost tracks; reference sample; systematics due to time dependence.
    • Write some text about ghosts ASAP!! (Sergey)

DC92 Release Report (March 2011)

  • Color code
    • Red: Obligatory for release
    • Green: Done
    • Black: Desirable, but lower priority

Recoil

  • Chi2-cut: <13.7. agreed for release. "conservative approach."
  • rdHypo selection algorithm & rdPID cut.
    • Agreed algorithm for release: select rdHypo w/o rdPID cut
    • Study of rdPID cut
      • Momentum- and quadrant dependence. Jen Studies shown 16.2.2011.
      • Apply optimized rdPID cut (i.e. rejecting pions) a) before kinematic fitting, b) after kinematic fitting. Jen Conclusion 7.3.2011: a) and b) give no or very little difference. Moreover, the extracted pure elastic asymmetries are very stable under an rdPID cut (i.e. no cut compared to cuts of varying strictness).
  • rdMultiplicity
  • Recoil inefficiencies.
    • Include Recoil inefficiencies into MC (XTC level). Efficiency maps exist for 2006 and 2007 (integrated) binned in momentum and fiber number. done by Sergey for the SFT, MC production exists.
      • Study impact of inefficiencies on angular distributions and amplitudes. Extract amplitudes from MC with inefficiencies switched off and on (same generated events). Their difference can be assigned as sys. uncertainty. Sergey 7.3.2011: Differences in central asymmetry values are very small.
  • Misalignment studies involving the Recoil detector.
    • Extract amplitudes in each quadrant separately, if possible binned in kinematics. Is there a difference? Could be assigned as systematics. Caro's studies on BSA quadrant dependence Sergey, Caro Shown to result in no systematic effect.
  • Purities and efficiencies from a combined gmcDVCS/DISng Monte Carlo.
    • Purities and efficiency for pure elastic sample vs. upper chi2-cut. Caro
    • Plot fractions of elastic and associated vs. t for the traditional DVCS sample, the reference sample, and the pure elastic sample. Caro
    • Calculate estimated associated asymmetry from reference sample, pure elastic sample and MC fractions. Sergey and Morgan
  • Final optimization of kinematic fitting.
    • kinfit(1) version as of Jan 25, 2011. Results of amplitudes shown on Feb 2, 2011.
  • Reference ("Master") sample.
    • The reference sample contains events as determined from a traditional DVCS analysis using only the front spectrometer, however only those events in the kinematic phase space for which the Recoil proton WOULD BE in the Recoil acceptance. Without use of Recoil information.. In the meeting of 16.2.2011, the importance of having such a reference sample was pointed out: The reference will demonstrate that the larger magnitude of the pure elastic amplitude is not (only) due to the different kinematic phase space, but that the pure elastic amplitude really presents a physics result.
    • Use Dietmar's acceptance function. Caro
    • Use Andreas' advanced acceptance parameterization. Sergey
    • Caro's developing studies Caro, Sergey
    • How can we estimate a systematic uncertainty due to the acceptance function?

Spectrometer

  • Study HTC track probability cut. Sergey
    • The asymmetry amplitudes for the standard cut of 0.01 and 0.0001 were compared and found to be almost identical. This systematic study is considered to be closed. 23.2.2011
  • Use new z-distance of calorimeter (as of Feb,2011) for the productions with new alignment: z=734cm
    • Both for photons and tracks.
  • Low- and High-density running.
    • Study if there is any difference in the asymmetry amplitudes results. Jennifer Done and shown 16.2.2011 to be consistent.
  • H0 inefficiencies.
    • Approach: extract amplitudes from an analysis that does not use tr21 (default), but tr18__tr26. If there is no change, close the subject. No difference found. Shown Feb 2 Caro

Systematic uncertainties

  • Time dependence of exclusive yield.
    • Determine systematics.
    • Impact of applying Charlotte's way of correcting the photon energy per time period. How does it influence the cross-check constant amplitude of the BSA? Sergey Done. Shown to have no effect 16.2.2011.
    • Mx2 time dependence with chi2<CUT cut applied. Caro shown Jan 25, 2011. No change in trend of time dependence.
  • Traditional uncertainties. Glasgow crew
    • "all-in-one": first results by Jen are available, as included in the extended analysis report and shown on 16.2.2011.
    • Remaining fractions of associated, SIDIS and exclusive neutral mesons? Uncertainty due to that? Exclusive pi0 are not expected to completely be killed by kinematic fitting. It can happen that both pi0 photons end up in the same calorimeter cluster.
  • Recoil inefficiencies: see above.
  • Misalignment involving Recoil: see above.

Related Pages

Abgerufen von „https://hermes-wiki.desy.de/index.php?title=DVCS_with_Recoil_Analysis_Page_DUMP&oldid=13283“