Oct 10/25 PionLT/KaonLT Analysis Meeting Notes
                ----------------------------------------------
                                (Notes by GH)

                    Today: KaonLT will be discussed first

Please remember to post your slides at:
https://redmine.jlab.org/projects/kltexp/wiki/Kaon_LT_Meetings

Present
-------
Regina - Garth Huber, Alicia Postuma, Nathan Heinrich, Nermin Sadoun,
   Vijay Kumar, Nacer Hamdi, Muhammad Junaid
FIU - Pete Markowitz
CUA - Chi Kin Tam, BC Manoj, Tanja Horn, Sameer Jain
Virginia - Richard Trotta
Ohio - Julie Roche
CSULA - Konrad Aniol
JMU - Ioana Niculescu, Gabriel Niculescu

Vijay
-----
Summer 2019 PionLT data analysis
- CoinBlocking correction
  - made plots of CTime.CoinTime_RAW_ROC1,2
    - ROC1 plot is cleaner than ROC2, will use this one
  - found oscilloscope traces of trigger timing at logentry #3655655
    - based on oscilloscope tracs, the timing cut window should be ~140ns wide

- pion absoprtion correction
  - will apply correction previously determined with Alicia's Geant4 simulation

Sameer
------
KaonLT high Q2 CoinBlocking correction
- looking at Q2=2.115 data taken on 2018.10.02

Chi Kin
-------
KaonLT Q2=3.0, W=3.14 data analysis
- MissingMass checks
  - applying cut 1.10<MM<1.14 to select K+Lambda region for both data and MC
  - pion subtraction applied to data
  - previous error needing large MM shift to data was fixed, the shift is only
    a few MeV now.  However, correcting both data and MC, as neither are at PDG
    mass
    - *NB* Tanja: you need to remove the MM shift from the MC, the Lambda peak
      will be at a mass slightly higher than PDG value due to energy loss in
      the target and other materials.  Just shift the data to the MC peak
      position

- Data & SIMC plot comparisons
  - various differences in the distributions, particularly for right SHMS
    setting
  - Gabriel: pion leakthrough subtraction has a different delta-distribution
    than K+Lambda data, how do you correct for this?
    - Richard: shows Q2=4.4 pion subtraction
      - needs to add an empirical fit underneath the Lambda, Sigma peaks that
        is not removed by the pion leakthrough subtraction
      - RF timing information was added mid-way through Q2=3.0, W=3.14 setting.
        The Q2=5.5 setting has it, so these data have much less pion
        leakthrough than the other settings.
  - Tanja: Chi Kin is intentionally treating the background different than
    Richard, so we can better understand how reproducible the results are

- LT-sep iterations with 10 phi-bins
  - after 4 iterations, Data/MC ratios are much closer to unity
  - L/T/LT/TT functional forms are the same as Richard's, but parameters are
    different
  - separated cross section ratios are fairly stable from IT=0,4,8
  - Julie: the Data/MC ratios have phi-dependence, but the sig_uns are flatter.
    Does this mean the model has a larger (and wrong) phi-dependence compared
    to the data?
    - Tanja, Garth: yes this is what it means.  The LT/TT cross sections are
      determined directly from the Rosenbluth fits of sig_uns.  It will be
      important to confirm the model dependence in the result is small.

Alicia
------
Difficulties re-submitting the pi+n BSA paper to PhysLettB
- editorial manager portal has changed completely since initial submission
  - need to upload XML info for all authors via undefined "standard procedure"
    - no guidelines given, following INSPIRE-HEP instructions for now
    - using ror.org Research Organization Registration for institution
      addresses, which is helpful
  - need to include ORCiD numbers for all authors included in MetaData.  Did a
    lot of work to find ORCiD for most people, but a few have left the field and
    don't have one
 - Tanja: please contact Haiyan Gao directly, including screenshots of where
   you're encountering problems

Richard
-------
Computer Physics Communications paper on LT-function fitting algorithm
- combined Fpi-2 data for both Q2=1.6, 2.45 into one dataset (t-dep only) and
  put it through the fitting algorithm
  - algorithm improves the Chi-squared compared to earlier results
- got CSV files from Vijay(pi+n) and Nacer(K+L) on their separated
  low Q2 cross-sections
  - next step will be to include these in the fits

Q2=4.4 LT-separations
- data has pion subtraction and empirical fit underneath Lambda
  - no Sigma fit is included
  - equations slightly different than what Chi Kin is using
- Data & SIMC plot comparisons
  - center SHMS setting, SIMC has narrower Q2,W-distributions than data
  - acceptance plots look good for both SHMS settings
- Data/MC ratios between 0.75 and 1.5
  - surprisingly, the lower -t bins have worse ratios than high -t bins
    - might be a pion leakage issue at low -t, Richard will have to look into
      it
    - *NB* Tanja: please make a comparison of Richard and Chi Kin's
      background subtractions for this setting
- made a weighted verage of phi-bins at each epsilon similar to what Vijay
  showed last week
  - it shows that high epsilon>low epsilon, which is good news
  - sigL<<sigT for all bins
  - Garth and Tanja willing to compare separated results with VGL model when
    things are ready

Nacer
-----
KaonLT Q2=0.50 LT-separations
- rechecking Data & SIMC MM distributions to be sure pion leakthrough is not an
  issue
  - to recap, sigL has a surprisingly large peak at low -t that is suspicious
  - RF timing cut is available for these data, so pion contamination is small
  - Data/MC ratios are fairly flat 0.8-1.2 at low epsilon, some phi-oscillations
    at high epsilon, particularly for higher -t bins
  - shows MM plots for 7 t-bins for center SHMS low epsilon, then shows for
    t-phi bins (i.e. dividing each t-bin into phi bins)
    - pion subtraction looks very good, no evidence that background is the
      cause of the weird sigL at low -t
    - *NB* Garth: SIMC normalization seems off for t-phi bins, compared to
      t-bins, plotting script needs to be rechecked

- will go back to model optimization
  - had LT=TT=0 functional forms in model
  - earlier, had tried Fpi-2 parameterization, but this gave bad ratios, so
    stopped
  - then tried changing LT,TT functional forms to same as L,T forms in
    sig_factorized_2007.  This looks more promising, so will do more iterations

Nathan
------
PionLT Coin Lumi analysis
- calculating systematic uncertainties on Coin Blocking correction
  - will move timing cuts by 5 ns in and out
  - *NB* Vijay: suggest to take 1 beam bucket (4ns) left and right instead
  - will take symmetric cuts around main distribution for systematic
    uncertainties
    - Garth: unless there is a good reason for asymmetric cuts in the error
      analysis, suggest to keep the cuts as symmetric
  
- calculating LiveTime systematic errors
  - compared CPU*ELLT to EDTM values
  - the difference gave a ~4% difference at high coin rate, which is
    suspiciously similar to the AdHoc correction discussed earlier
    - Garth had suggested to recalculate the normalized coin yields (after Coin
      Blocking) with EDTM instead of CPU*ELLT
      ** the resulting distribution is flat, which means no AdHoc correction
      ** would be needed
    - clearly this means the EDTM is capturing some element of deadtime that
      CPU*ELLT misses
      ** will use EDTM for all coin data, and no AdHoc correction!
      - EDTM is still unreliable for singled data though
- will need to evaluate a systematic uncertainty in the EDTM
  - suggests to fit each TLT vs Rate for each CoinLumi setting and look at the
    spread of points from the fit curve
  - concerned that simply combining the variations between the settings, as the
    TLT differences them (at same CoinRate) are probably real
  - *NB* Gabriel: suggests to make a histogram of the residuals between the
    data and fit function and see if this gives a Gaussian distribution
  - *NB* Julie: the used logarithmic fit diverges at zero rate.  It's important
    to switch to an exponential fit that goes to 1 at zero rate

Junaid
------
PionLT Q2=3.85 LT-separations
- working on LT,TT functional forms to reduce oscillations in Data/MC ratios
  - ratios are 0.8 to 1.2, but with 2-bump oscillations coming from TT
  - Rosenbluth fits give sigL>sigT at low -t and fairly small LT,TT
  - will need to modify the sigTT functional form
  - sigLT functional form seems fine, based on much smaller 1-bump oscillation
    in ratios

Nermin
------
PionLT data analysis
- met w/ Junaid on hcana software setup
- reading up on Hall C detector systems

Next Meeting
------------
We will try our new 2-meeting/week structure
- Thurs: Oct 16 @ 16:00 Eastern/14:00 Regina
  - PionLT will go first
- Fri: Oct 17 @ 11:00 Eastern/9:00 Regina
  - KaonLT will go first