Jet Tagging Scale Factors Measurement

Scale factor method

The scale factors calculated with this repository are designed to calibrate the shape of b-tagging discriminants (currently DeepCSV and DeepJet) for heavy flavor and light flavor jets.

Implementation Info

The Luigi Analysis Workflow (law) package is used to structure the scale factor measurement. It is based on the luigi pipelining tool, formerly developed by Spotify. Logical units of the measurement code are structured in so-called Task classes. Tasks produce output Target's and can require each other. This defines the actual workflow to run. If a task is run from the command-line, all required tasks whose outputs do not exist yet are triggered as well. See the links below for more information on luigi.

law is an extension ontop of luigi. It adds useful classes to work with infrastructure available to the HEP community. E.g. it supports submission to the WLCG as well HTCondor and LSF batch systems, and also storage of files on remote resources using a bunch of protocols (EOS, xRootD, dCache, GridFTP, WebDav, Dropbox, CERNBox, etc.).

Useful links:

• luigi: Repo, Docs, hello_world.py
• law: Repo, Docs (in progress), Examples

Setup

# source the main setup file
scram_cores=4 source setup.sh

# create the law task index file
# (required for bash/zsh autocompletion)
law index

# install CMSSW modules once
# (only needed when creating new trees)
law run InstallCMSSWCode

To let the tasks talk to a central luigi scheduler, you need to set $JTSF_SCHEDULER_HOST.

Running Tasks

Check the wiki page of this repository for instructions on how to run the tasks required for the SF calculations.