Warning, /HLTriggerOffline/Egamma/doc/README is written in an unsupported language. File is not indexed.
0001 ************************************************
0002 * Validation/Optimization Tools for HLTOffline *
0003 * by Joshua Berger *
0004 ************************************************
0005
0006 Instructions for running code:
0007
0008 1. Compile the code as usual using scramv1 b.
0009
0010 2. Find a suitable sample to run on. The sample should have RAW since the HLT now works from RAW.
0011 Insert this in PoolSource.fileNames in test/HLTRecoWithoutFilters.cfg.
0012
0013 3. Change the output file name to something suitable for the input file. The output of this stage
0014 will be a root file containing all the HLT isolation variables and some other quantities of
0015 interest for e/gamma. Then do
0016
0017 > cmsRun HLTRecoWithoutFilters.cfg
0018
0019 If you want to run on larger samples, you will probably want to run on a batch or on the grid.
0020 I have several sample scripts in ../test/HLTStudyData/ for running on the batches at FNAL and CERN as well as using CRAB.
0021 Getting grid credentials... now that's a different story.
0022
0023 For Validation:
0024 4. Edit ../macros/Get*Eff.C, where * is the name of one of the Egamma HLT paths: SingleElectron,
0025 RelaxedSingleElectron, etc. Change the TFile argument to the file created in 3. Change the cuts
0026 at the beginning of the file to the cuts you would like to test. These are currently the cuts in
0027 the actual HLT code as of 1_6_0. This should probably be updated to follow the newer format (files
0028 should be consolidated and user-defined multi-cut functions should be eliminated).
0029
0030 5. Do the following:
0031
0032 > cd ../macros
0033 > root.exe
0034 [] .L CopyTreeMulti.C; // Only if running a Double trigger
0035 [] .x Get*Eff.C
0036
0037 The output will be as follows: a text output to stdout of the number of events passing each filter
0038 and the efficiency of each filter and 3 graphs of efficiency vs. Et, eta, phi to
0039 macros/images/EffV(Et/eta/phi)(SE/RSE/etc.).gif.
0040
0041 Bugs:
0042 - Root scripts produce a long series of errors that do no affect the output. These will probably be eliminated
0043 if the code is updated using the new SUM$ function described in other files.
0044 - When eff = 1, the errors are calculated using some Bayesian algorithm in this version root. This is
0045 changed to the more usual "error = 0 for eff = 1" way of working in newer versions of root (that we are not
0046 using for some reason).
0047
0048 For Rates, Plots, Optimization, ...:
0049 6. Most of the files for these functions are documented with comments. They are all located in ../macros with
0050 self-explanatory (I think) names. Old Vars refers to files meant to study IHcal, E/p, while New Vars refers to
0051 files meant to study H/E, |1/E - 1/p|. Note that I have added H/E and |1/E - 1/p| to the output of the
0052 CMSSW part of the code. However, I have not updated the root scripts yet because not all the background samples
0053 contain this update! More details in the actual files where relevant. In order to run any of these scripts, do
0054 the following in the macros directory:
0055
0056 > root.exe
0057 [] .L <Script-Name>.C++
0058 [] <Script-Name>
0059
0060 Warning: the output is generally to file names defined in the script file.
