-- VivekJain - 27 Nov 2006

Some ideas for writing the Physics Analysis Work Book


Here are some thoughts about the PhysicsAnalysisWorkBook

  • We should structure it in a way that mirrors a person’s thought process when they start to think about doing analysis
  • We should provide a lot of examples, both simple and complex, with annotated python scripts, log files, source files, etc.
    • AnalysisSkeleton.cxx is a good file to orient people
    • Other examples can include topics like, how do I vertex a selected set of tracks, etc
    • There could be links to other standard software, e.g., RooFit (a layer on top of ROOT to do things like multi-variate likelihood fits), neural nets, etc.
  • We should have a list of people who are willing to answer questions in a timely manner, e.g., analysis tools experts, physics group conveners, etc. In addition, also provide a FAQ
  • ...

How do I run on data (or MC)

  • What formats are available? Point out the pros/cons of each format.
    • For instance, one can run on AOD’s: Pro – most pristine format with no cuts on objects, Con - this will require working in the Athena framework.
    • Root-tuples (whether AthenaAware or not): Pro – no knowledge of Athena framework required, Con –were cuts made to produce them?
    • Is there a preferred format? Point to tools like EventView or whatever other such tools might exist
    • ...

  • What information is stored in these formats?
    • What objects are stored? What variables for each of the objects? An explanation will be very useful.

  • Examples on how to run on the formats described above. Here they can just run interactively on a handful of events and look at the output.

  • How do I look at some event displays?

  • Here we can have more sophisticated tools, e.g., EventView or whatever comes next. We should encourage authors of these tools to have good documentation, e.g., if they introduce new variables, then there should be a good explanation of them

  • How do I include trigger information in my analysis? What triggers are available?

  • How do I run on a large(r) sample of data (or MC)?
    • Where do I run batch jobs? What do I need to do before I can do this, e.g., do I need a grid certificate, accounts on the "nearest" Tier 1 and/or Tier 2 centers?
    • Provide sample batch scripts (see Jesse Ernst's script to split jobs and submit them via condor at BNL)
    • Alternatively, one can submit jobs to run on the Grid. Has this been tested? Does it work?

Data Samples

  • Where do I find data samples? We should distinguish between small test samples and large scale running.
    • Explain how to find test samples. Should people use dq2_ls?
    • I tried using the dq2 dataset browser and I found many datasets were empty. In cases where these datasets did contain files, I got some cryptic error messages – this needs better documentation
      • Someone remarked to me, “The Panda Monitor” has a high degree of color coding, but sometimes it is not clear what things mean.
  • What about running on large data samples?
    • Is it better that people first get in touch with physics conveners or should they go off on their own?

MC samples

  • How do I generate small test samples
    • Very useful to produce Evgen files to ensure that I am setting up things correctly. How do I run on these files?
    • Some people may need to filter events when generating events. It seems that there is no centrally approved software to do this and that everyone is writing their own filters?
      • Is that what is envisioned?
      • Can MC production farms deal with these filters?
    • Produce AOD’s (or the preferred format) – already explained how to do this in the standard Work Book or above

  • What about large samples? How do I go about getting a large sample of events? Are there already there? Perhaps it is best to go through physics conveners.

Edit | Attach | Watch | Print version | History: r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r3 - 2006-11-29 - VivekJain
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback