CAFAna-for-cross-sections workshop

Phone connection (866) 740 1260 2236837#

Thursday, August 11th

 9:00 - 10:30  Curia II (WH2SW)
   Intro/CAFAna basics (Chris)
   Flux/Target Count/Efficiency definitions (Chris)
   Truth matching and nuTree

10:30 - 11:00  break

11:00 - 12:30  Curia II (WH2SW)
    Toy analysis (Dan)

12:30 -  1:30  lunch

 1:30 -  3:00  Race Track (WH7XO)
    Friendly interface (Dan)

 3:00 -  3:30  break

 3:30 -  5:00  Conjectorium (WH3NE)
   Overflow and hacking time (toy analysis)

Friday, August 12th

 9:00 - 10:30  Aquarium (WH15SW)
   numu inclusive (Kanika)
   nue inclusive (Pengfei)
   CC pi0 (Dan)
   NC COH (Duyang)

10:30 - 11:00  break

11:00 - 12:30  Aquarium (WH15SW)
   EnsembleSpectrum (Jeremy)
   Sidebands (Jeremy)
   Cross-section ratios

12:30 -  1:30  lunch

 1:30 -  3:00  Aquarium (WH15SW)
    Reusing constraints in xsec analyses (Jeremy)
    Particle-level MCTruth (Kanika)
    Wishlist items (all)

 3:00 -  3:30  break

 3:30 -  5:00  Aquarium (WH15SW)
    Overflow and hacking time (real analyses)

This agenda is a sketch only. If we need more or less time for a particular topic that's fine.

Preliminary work

Chris: make it possible to fill a spectrum from the nuTree (done, untested)


The main focus will be on discussion of the tools needed for rigorous ND analyses, what exists in CAFAna, and what may need to be developed.

I expect a lot of discussion, but having a handful of analyses (whether or not their physics is in a final state) that can try and work through the whole chain in practice would be useful.

Here are some more slightly more detailed notes on the topics above:

  • Flux
    - I've previously showed a method for deriving the neutrino flux from the CAFs
    - This is all tangled up with questions about target counts/fiducial volumes/efficiencies
    - It also for sure needs us to store truth information for unsliced neutrinos (nuTree)
  • Target count
    - There's code for this, recently upgraded by Kanika to be able to apply a fiducial volume
    - Tangled up with definitions of efficiency etc again.
  • Efficiency
    - See above. I owe people an attempt at a mathematical write-up so we can see that all necessary factors are included (or not)
    - This is also connected to the question of how exactly we do truth matching
  • Unfolding
    - Multiple techniques exist. Need hands-on experience
    - Some advances might be necessary for unfolding to 2D distributions, or using a sideband as a constraint
  • Systematics
    - Various tools exist for oscillation analyses
    - Do xsecs have any additional needs?
    - How best to carry uncertainties through the whole chain?
    - Is "SpectrumInterp" useful/necessary? Do we need to be able to break down by non-flavour categories?
    - Publications expect is to produce a correlation matrix? From an ensemble of random throws?
  • Sidebands
    - How do they work?
    - Constraining a systematic: how does that work if the signal is in principle completely unknown?
    - Relationship to unfolding?
  • Cross-section ratios
    - Anything special to consider here? Have to ensure systematics cancel correctly
  • User-friendly wrapper
    - It should almost be possible to make a class where one puts in the signal channel definition and selection cuts and a cross-section with error band comes out. Bringing such a thing closer to reality is one goal of the workshop
    - Let's start without any systematic errors
  • Prong-level truth (Kanika)
    - Useful for development of reco/PID algorithms
    - Associated with in-progress CAF rewrite (Kirk?)
  • ND constraints on oscillation analyses
    - Not strictly related to obtaining xsec measurements, but we should think about how best to feed such measurements back into the oscillation fits (Jeremy)
  • Special needs of particular analyses
    - Do you think your analysis is special, needs something not listed above? We'd love to hear from you
    - Representative analyses: NC COH (Duyang), numu incl. (Kanika), nue incl. (Pengfei), CC pi0 (Dan),

Remote attendees: Jeremy, Travis, Shih-Kai