(r7) WJetsAnalysis < Main

Tags: , view all tags
---+ W + Jets Analysis

   * First, set up CMSSW 1_6_0 as described in CMSSWinSprace

   * Download the tar.gz file that contains the relevant files, and unpack it somewhere. It automatically creates the =analysis_SPRACE= directory.

   * Create an EDAnalyzer called !VnjetAnalyzer (the name doesn't really matter, it's just to keep source file compatibility). From the =$CMSSW_BASE= dir (the =CMSSW_1_6_0= dir in this case), issue the following commands:

<verbatim>
$ mkdir VnjetAnalyzer
$ cd VnjetAnalyzer
$ mkedanlzr VnjetAnalyzer
</verbatim>

This creates the following directory hierarchy:

<center>
<img src="http://www.sprace.org.br/Twiki/Images/VnjetAnalyzer_hyerarchy.png" alt="Fig.1"/>
</center>

   * Copy the EDAnalyzer (in our case, the =VnjetAnalyzer.cc= source file) from =analysis_SPRACE= to the =src= directory - overwriting the one you just created. Study the structure of the file, and how it uses the Analysis Tools. For more information, check the [[https://twiki.cern.ch/twiki/bin/view/CMS/WorkBookIntroAnalTools][CMS Workbook in Analysis Tools]]. Also, copy the =BuildFile= to the (deeper) =VnjetAnalyzer= directory.

   * Issue the =scramv1 b -r= command to compile the EDAnalyzer - it should compile fine. If it does not, you probably misplaced the =BuildFile=. Correct the mistake and issue the command again. The =-r= option tells the =scramv1 b= command to reread all files - if the =BuildFile=s have not changed, it is usually safe to skip that option.

   * Copy the *Configuration Fragments* =w0.cff=, =w1.cff=, ..., to the =$CMSSW_BASE/share= directory. These Configuration Fragments contain the location of the sources (ROOT files) for our analysis. The sources contain w + jets Events. The number following the letter w is the number of hard partons that were simulated at matrix-element level. After that, the final state is subject to parton showering and hadronization. These files are NOT used at the moment, but they may be used in the future if the samples are stored in the SPRACE cluster at some point in the future.

   * Copy the *Configuration Files* =w0.cfg=, =w1.cfg= to the =VnjetAnalyzer= directory. These Configuration Files accomplish the following tasks: scan through each Event in the source, select all partons, create parton-level jets ("clusters") with a given cone radius (default is 0.5), cut on the pt of the jets (default is 20 !GeV), and produce a series of histograms. Let's focus our attention in two particular histograms: "totalJets", with the number of jets per Event, and "pt_tJets" with the cumulative pt distribution of jets for all Events. Each .cfg is linked to one of the .cff, and produces a ROOT file named =wnj_ic_pt20_R05.root=, with n the same as in the .cfg and the .cff.

   * Copy this ROOT macro: =group.C= to the main directory. What this macro does is to group all the ROOT files produced by the .cfg files in a single ROOT file. It also sums the histograms for number of jets per Event and cumulative pt distribution in both exclusive and inclusive ways. For instance, in the "nexclusive" histogram which is created, each bin corresponds to the number of Events that have EXACTLY that number of jets. (There is also a luminosity scale factor involved, so that for a given .cfg file the "totalJets" histogram corresponds to 100/pb). The "ninclusive" histogram works in a similar fashion, except that each bin corresponds to the number of Events that have AT LEAST that number of jets.

   * Run all the Configuration Files: =cmsRun w0.cfg=, ... This will produce the =wnj_ic_pt20_R05.root= files.

   * Run the ROOT macro: =root -l -q group.C=. This will produce the "allw_ic_pt20_R05.root" file.

   * Plot the resulting file with ROOT: =root -l allw_ic_pt20_R05.root=, and select the "nexclusive" histogram. Plot it in log scale to see that the log of number of events with n jets is a linear function of n. That is because the cross-section for n jets scales as =a^n=, where =a= is the strong coupling constant, ("alpha_strong"). If one now fits a linear function to that histogram, the slope should be =ln a=. It works similarly for the "ninclusive" histogram.

-- Main.ThiagoTomei - 16 Nov 2007