GlueX Analysis Workfest 2018

From GlueXWiki
Revision as of 16:17, 17 May 2018 by Jrsteven (Talk | contribs) (Thursday May 17, 2018)

Jump to: navigation, search


  1. To discuss analysis tools, techniques, and results required as input for amplitude analyses and absolute cross section measurements.
  2. To actively work in groups on solving these problems, with well defined deliverables to come from the Workfest.

Workfest Tasks

The workfest tasks are split into two teams below. Team mascots and logos are up for debate... suggestions welcome.

R&D Team Tasks (Justin)

1) Comparison of track and shower efficiencies between data and MC (Simon, Alex A., Jon, Will M., Ahmed)

  • Suggested studies: Missing particle in well-defined final states and measure efficiency of reconstruction
    • Possible reactions for tracking effic.: Missing π or proton in γp → π+π-p, γp → π+π-π0p, γp → π+π-π+π-p
    • Possible reactions for shower effic.: Missing γ in γp → π+π-π0p (FCAL), γp → π+π-π0π0p (BCAL?)
      • Previous studies: Jon Zarling [4], Cristiano Fanelli [5] [6] [7], Jake Bennett at BESIII [8]
  • Deliverable: Initial evaluation of efficiency vs kinematic variables (p, θ, φ) for data and MC
  • Bonus points: Initial evaluation of resolution (measured - missing particle KinFit) for kinematic variables (p, θ, φ)

2) Comparison of kinematic fit χ2 between data and MC (Daniel, Alex B, Stuart, Mike M.)

  • Suggested studies: Analysis of multiple final states with different constrains (P4, P4+vertex, P4+vertex with mass constraints)
    • Possible reactions: γp → π+π-p (ρ and Δ++), γp → K+K-p (φ and Λ(1520)), γp → π+π-π0p (ω and η), γp → π0γp (ω), γp → π+π-ηp (η'), γp → π0π0ηp (η')
    • Previous studies: Alex Barnes [9] Daniel Lersch [10] [11] Pre-Studies
  • Deliverable: background subtracted χ2 and pull distributions for data and MC for charged, neutral, and mixed final states Day 1Day 2
  • Bonus points: Prepare for tuning of MC parameters and/or covariance matrices to improve data/MC agreement

Validation Team Tasks (Sean)

1) Comparison of branching ratios between data and MC (Mark D., Tegan, Mahmoud)

  • Suggested studies: Analysis of multiple decay modes for ω, η, and η' to measure ratio of branching ratios (with MC efficiency correction)
    • Possible reactions: γp → γγp (η), γp → π+π-π0p (ω and η), γp → π0π0π0p (η), γp → π0γp (ω), γp → π+π-ηp (η'), γp → π0π0ηp (η')
  • Deliverable: background subtracted and efficiency corrected ω, η and η' yields in different decay modes
  • Bonus points: investigate dependence on kinematic variables: Eγ, -t, etc. and including/excluding specific detector regions

2) Cross section evaluation for benchmark channel (Thomas, Simon)

  • summary
  • Suggested studies: Evaluate all ingredients of cross section: yield, efficiency, flux for benchmark channel
    • Possible reactions: γp → π+π-p
  • Deliverable: scripts to evaluate cross section for benchmark channel with a given version of software/calibrations
  • Bonus points: implement in monitoring database

3) Event-by-event studies of benchmark channels (Sean, Ashley, Lubomir)

  • Suggested studies: determine yields for pre-skimmed datasets of rare channels
    • Possible reactions: J/ψ, Ξ, Δ++
  • Deliverable: scripts to monitor event yield and/or other observables with a given version of software/calibrations
  • Bonus points: implement in monitoring database
  • Summary 1 (Sean)
  • Summary 2 (Lubomir)

4) Quantitative comparison of data and weighted phasespace MC for angular dependent analyses (Alex B., Elton, Matt, Carlos)

  • Suggested studies: Fit vector meson SDME using current MC efficiency to compare weighted MC with data
    • Possible reactions: γp → π+π-p (ρ), γp → K+K-p (φ)
  • Deliverable: quantitative comparison (χ2, Kolmogorov-Smirnov, etc.) of weighted phasespace MC angular distribution with data
  • Bonus points: account for background (accidentals, etc.) in AmpTools, and statistical tests of fit parameters (toy MC, bootstrap, etc.)
  • Summary

Workfest Software and Conditions

Please use the branches specified below when checking in changes for workfest activities.

  • version.xml: /group/halld/www/halldweb/html/dist/version_workfest2018_baseline.xml
  • sim-recon branch: workfest2018
  • hdds branch: workfest2018
  • gluex_root_analysis branch: workfest2018

For generating and analyzing MC we'll use a special variation and timestamp as the default:

  • calibration time: calibtime=2018-05-08-08-00-00
  • calibration variation: mc_workfest2018

GitHub repository for workfest scripts

Analysis scripts developed for tasks should be stored in the gluex_workshops/workfest_2018 GitHub repository. To get the default environment on the ifarm (see above), please use this setup.csh script in the repository.

Workfest Datasets

2017 Low Intensity Data

  • 3 days at the end of Spring 2017 Low Intensity totaling ~4.5B events (~10% of the Spring 2017 data)
  • Run numbers: 30730-30788
  • EVIO location: /cache/halld/RunPeriod-2017-01/rawdata/Run0307??/
  • REST location: /cache/halld/offline_monitoring/RunPeriod-2017-01/ver33/REST/0307??/
  • Analysis TTree location:
    • per file: /volatile/halld/analysis/RunPeriod-2017-01/ver12/
    • merged: /cache/halld/RunPeriod-2017-01/analysis/ver12/

MC Simulation Samples


  • Run number: 30730
  • Background: random trigger
  • Beam energy range: Eγ = 3 - 11.6 GeV, with coherent edge set at 8.8 GeV


Generator Number of events (106) Settings Planned usage
bggen 10 General inclusive sample for many studies
gen_etaRegge 1 η → γγ Branching Ratios, KinFit χ2
gen_etaRegge 1 η → π+π-π0 Branching Ratios, KinFit χ2
gen_etaRegge 1 η → π0π0π0 Branching Ratios, KinFit χ2
gen_etaRegge 1 η → π+π-γ Branching Ratios
gen_etaRegge 1 η' → π+π-η Branching Ratios, KinFit χ2
gen_etaRegge 1 η' → π0π0η Branching Ratios, KinFit χ2
gen_etaRegge 1 η' → π+π-γ Branching Ratios
gen_2pi_amp 2.5 Physical distribution for ρ → π+π- and Phasespace KinFit χ2, Efficiencies, Cross section, ρ SDME
gen_omega_3pi 2.5 ω → π+π+π0 Branching Ratios, KinFit χ2, Efficiencies
gen_omega_radiative 2.5 ω → π0γ Branching Ratios, KinFit χ2
gen_2k 2.5 Physical distribution for φ → K+K- and Phasespace KinFit χ2, Efficiencies, Cross section, φ SDME
gen_amp 2.5 4π Phasespace for tracking efficiency Tracking Efficiencies

Data location

  • HDGeant location:
  • Smeared HDGeant location:
  • REST location: /cache/halld/workshops/workfest2018/recon/ver02/generator_name/hddm/
  • Analysis TTree location: /cache/halld/workshops/workfest2018/analysis/ver02/generator_name/
  • MC thrown TTree location: /cache/halld/workshops/workfest2018/analysis/ver02/generator_name/tree_thrown/


Example starting framework:

Expected Outputs

Each validation package should generate the following:

  • Figures in PNG format for monitoring the output
  • Text file containing summary numerical values

The package should also provide

  • Description of each of the summary values
  • JSON formatted configuration options for the Dashboard page (description)

Experts (Sean, Thomas) will be in charge of displaying and storing this information.

Example Websites

Location and Time

The "Workfest" will take place at:

DATES: May 15 - 17, 2018



Please add your name to the list of attendees below and indicate which task(s) you're interested in contributing to. No formal registration or registration fee is required.

Name Participate at JLab (Yes/No) Task you're interested in contributing to
Justin Stevens Yes Track/shower efficiency and accidental subtraction/flux normalization
Simon Taylor Yes Track/shower efficiency and tuning covariance matrices
Sean Dobbs Yes Track/shower efficiency and general MC tuning, statistical tools
Alex Austregesilo Yes Track/shower efficiency and accidental subtraction/flux normalization
Daniel Lersch Yes Tuning the covariance matrix for kinematic fits
Thomas Britton Yes MC tuning and accidental subtraction/flux normalization
Alex Barnes Yes Tuning covariance matrices and statistical tools
Jon Zarling Yes Track/shower efficiency with ω mesons. Tuning of inclusive MC?
Mike McCracken Prob. remote Tuning the covariance matrix for kinematic fits
Mahmoud Kamel Yes validating the track/shower efficiency determination from MC using the η/η' branching ratios
Mark Dalton Yes Neutral particle covariance matrices, neutral shower efficiency, Comparison of branching ratios, Cross section evaluation for ω→3π
Elton Smith Yes Non-linearity in BCAL due to SiPMs, accidental subtraction in AmpTools
Tegan Beattie Yes
Ahmed Foda Yes
Matt Shepherd Yes
Stuart Fegan Yes
Sebastian Cole Yes
Georgios Vasileiadis No
Carlos Salgado Yes(Tuesday) Accidentals correction.
Lubomir Pentchev Yes Event-by-event studies of J/ψ channel
Andrew Schick Yes Non-linearity in BCAL due to SiPMs, accidental subtraction in AmpTools'
Igor Strakovsky No
Colin Gleason Yes
Mark Ito Yes Software support

Remote Participation

Remote participation will be possible, but the intention is to have a group of people actively working together at JLab. Thus remote participation will not be optimal.

  1. To join via a Web Browser, go to the page [12]
  2. More information on connecting to bluejeans is available.


Tuesday May 15, 2018

Discussion summary:

  • Tracking efficiency: Alex and Simon have trackeff_missing working to evaluate track efficiencies following Paul's procedure
    • Suggestion to add ΛK channel for future efficiency studies due to narrow mass spectrum
  • Shower efficiency: Jon and Ahmed studying MC samples with random trigger background to investigate "tail" in omega recoil spectrum
  • KinFit studies: Preparing update to DEventWriterROOT which optionally outputs track Pull information in standard analysis TTree (testing ongoing)
  • Event-by-event studies:
    • Lubomir has a script to produce histograms and obtain yields for J/psi events. David has a fix for the issue of analyzing events from multiple runs in a single hd_root process
    • Sean is working on some similar scripts for Cascade yields
  • Angular dependent analyses: Matt and Elton are investigating a bug in AmpTools for weighting events in the fit

Wednesday May 16, 2018

  • Announcements:
    • Overnight analysis launch results appearing for data as \ver12 and simulations \ver02 (see earlier section of wiki page for full path)
      • Includes change to particle ordering in ReactionFilter and additional requested plugins
      • Also adding samples of 4π and η→π+π-γ
  • 9:00 Informal update on tasks and discussion (coffee provided) (60)
  • 10:00 Begin working session
  • Lunch on your own
  • 13:00 Coffee available
  • 15:00 Informal update on tasks and discussion (coffee provided)
  • Plan for dinner outing to County Grill

Discussion summary:

  • Dashboard: Some issues with Windows/Firefox, but framework ready for monitoring many yields in one system. b1pi and Delta++ implemented, more to come from the workfest. Request to correlate values from different tables/channels.
  • Angular dependent analyses (Elton): Matt fixed the weight issue in AmpTools which was in the ROOT event reader in sim-recon. Good agreement in the 2 methods of: negative weights for accidentals, or fitting with signal and background files in AmpTools. Future statistical tests of these 2 methods. Some discussion about analysis flow from DSelector to AmpTools, needs further discussion offline.
  • TOF matching (Beni): Showed good data/MC(bggen) agreement in matching efficiency, except near the beam hole. Further investigation of TOFPoint definition for single ended paddles necessary. Additional tests suggested: 1) study MC without random trigger hits and data from 2017 HI for comparison and 2) study subtraction accidentally matched hits.
  • Tracking efficiency (Alex A): Has Paul's framework running. Shared many plots of pi- "resolution" for data and MC. More info on efficiency to come tomorrow.
  • KinFit: Plugin checked in which writes Pulls to analysis TTree. Outlined list of channels to be investigated.

Thursday May 17, 2018

  • ROOM CHANGE A110 for morning discussion: 9:00 Informal update on tasks and discussion (coffee provided) (60)
  • F326 available starting at 11 am today, so afternoon discussions will be back there.
    • Neutral efficiency update (Jon)
  • 10:00 Begin working session
  • Lunch on your own
  • 13:00 Coffee available
  • 15:00 Summary of deliverables achieved and discussion of future work (coffee provided)
    • Discussion of Workfest format and future possibilities

Afternoon discussion summary

  • Tracking Resolution and Efficiency:
    • Simon: Missing tracks in pi+pi-p events from trackeff_missing plugin, uses some kinematic matching FOM between missing particle and reconstructed particle under study. With large statistical errors finds tracking efficiencies of ~80% in the data (less than 1 run). Need much more statistics and comparison to MC.
    • Alex A: Missing tracks in pi+pi-p and pi+pi-pi+pi-p events from trackeff_missing plugin, studying vs kinematic variables using Paul's scripts (without missing mass sidebands). No cuts on KinFit, so MM aren't biased. Comparisons of data and rho MC samples, providing efficiency in terms of p, theta, phi.
    • Action items:
      • For larger kinematic coverage suggestions to simulate bggen, Delta++pi- and Delta0pi+ sample for larger angle pions than rho sample provided.
      • Identify particular regions in phase space to compare check reconstruction quality before next REST production.
      • Investigate alternative method for track efficiencies, similar to shower efficiencies of reconstructing narrow peak (Lambda or omega) and fitting yields.
      • Summarize average efficiency (1D slices) for simpler comparison between data and MC
  • Shower Resolution and Efficiencies:
    • Colin: Photon missing in pi+pi-pi0p events, using similar selection as Jon Z. Request for more statistics in omega MC. Need to invest
    • Shower efficiencies from omega analysis discussed at morning meeting
    • Action items:
      • Large bggen and signal omega MC needed for shower efficiencies
  • Kinematic fit

Brainstorming on topics for future Workfests

  • Particle Identification (PID) cuts: definition in terms of CL, combine between different detectors, efficiency of cuts in data/MC, etc.
  • Implementation of tuning mcsmear parameters and covariance matrices
  • Statistical tools for validating fit parameter uncertainties in AmpTools