Minutes for 28 June 2011

1. Introductions, who's interested in what. (I've tried to summarize; feel free to edit your entry. An ", etc." is assumed after each entry.)
   1. Chris: Leader; software implementation of tools, automation of comparisons
   2. Jay: Working out benchmark signals, event generation
   3. Stefan: Event generation with Sherpa, comparing to other generators/shower models
   4. Lily: Two-body tagging
   5. Andrew: Detector effects
   6. Martin: Detector effects
   7. Miguel: Detector effects, Peter's simulator
   8. Marcel: same
   9. Kirill: not sure yet
   10. Peter Skands (not present): input on tunes, matching in Pythia
   11. David Miller (not present): pile-up?
2. Discussion of scope: What comparisons to make?
   1. Which signals? Probably at least top and W/Higgs/Other (WHO). Anything else?
      1. Definitely top and two-body (at least Higgs).
      2. Jay will come up with a few other benchmarks with more complicated signatures.
   2. What samples? That is, which MC issues to explore (shower models, matching, etc.)
      1. Baseline will be merged Sherpa samples
      2. Would like to compare to Pythia/Herwig 2->2 samples from last year (all pT bins??)
      3. Perhaps also include Madgraph+Pythia MLM matching vs. Sherpa's CKKW?
      4. Should consider tune effects in MG+Pythia; consult Peter on this.
   3. What effects? See laundry list above.
      1. Definitely consider detector effects. Miguel and Marcel have discussed this and will follow up, with further discussion in next week's meeting.
      2. Definitely consider pile-up somehow. No pile-up is totally unrealistic, but so is totally uncorrected pile-up. Need to decide what kind of initial pile-up removal is semi-realistic (just remove all charged pile-up particles, e.g.). Chris, Lily will follow up with David on this point.
   4. Which tools? Ditto.
      1. For tops, WHO, group seems happy with list above. Relatively clear which make sense for two-body signals.
      2. For other signals, will let Jay pick some benchmarks, then think more about which specific methods make sense.
   5. Another item not on the agenda: dijet correlations. We should take a look at mis-tag correlations in the background samples, these are essentially always assumed to be independent and it would be good to check. It would be good to compare generators here; there may be correlations between jets that each generator handles differently.
3. Practical considerations (should largely be decided by those who do the work, but worth discussing, I think). Didn't get here, so these are still decision points for next week.
   1. What is the comparison metric?
      1. Should at least do some with 2010 metric (BG eff. vs. signal eff.)
      2. How to compare before/after effect X? Show two 2010 plots, or do, e.g. before vs. after BG eff. plots? Someone should get out a copy of Tufte and think about this one.
   2. How to optimize consistently? Most methods have some input parameters, and some cut parameters (e.g., R_sub is an input, but m_W is usually a cut). This is closely related to the next item.
   3. How to implement tools? I'm working on getting all of this is SpartyJet, but in a way that's easy to pull out as standalone. Open to comments on how this should be done.
4. Time frame/plan going forward. Depends largely on scope!
   1. Who wants to really work on this (versus just provide input)? Didn't really discuss; let people take on what they want next week.
   2. Should this be a standalone paper? Not discussed; let's talk about this next week.
   3. Make assignments and plan for next meeting. Next meeting: same time next week. Assignments:
      1. Chris: Get tools implemented, follow up on pile-up questions with experimentalists
      2. Jay: Pick some benchmark samples
      3. Stefan: Start generating Sherpa samples
      4. Miguel/Marcel: Summarize work/discussion so far on detector issues, with more discussion planned for the next meeting.