btaggingEffAnalyzer

Make b-tagging efficiency maps with ANATools using datasets generated by HEPAnalysis.

How to use?

You can inspect the python notebook files as an example of how things are done, otherwise you can run the generate.py script specifying generation arguments. Example:

python generate.py \
    --period 16 \
    --basedir /dust/btaggingEffMaps \
    --algo DeepJet \
    --working-point medium \
    --eta-bins 0. 0.6 1.2 2.4 \
    --pt-min 20. \
    --pt-max 1000. \
    --pt-max-thr 0.001 \
    --step-size 5. \
    --unc-stop 0.3 \
    --unc-increase 0.0002 \
    --output-path ./output

For and APV dataset you can specify the --apv argument as following:

python generate.py \
    --period 16 \
    --basedir /dust/btaggingEffMaps \
    --apv \
    --algo DeepJet \
    --working-point medium \
    --eta-bins 0. 0.6 1.2 2.4 \
    --pt-min 20. \
    --pt-max 1000. \
    --pt-max-thr 0.001 \
    --step-size 5. \
    --unc-stop 0.3 \
    --unc-increase 0.0002 \
    --output-path ./output

It is also possible to limit the number of events that are going to be swept using --max-events argument as following:

python generate.py \
    --period 16 \
    --basedir /dust/btaggingEffMaps \
    --apv \
    --algo DeepJet \
    --working-point medium \
    --eta-bins 0. 0.6 1.2 2.4 \
    --pt-min 20. \
    --pt-max 1000. \
    --pt-max-thr 0.001 \
    --step-size 5. \
    --unc-stop 0.3 \
    --unc-increase 0.0002 \
    --output-path ./output \
    --max-events 2000

Using -1 will consider all available events in the dataset.

When specifying the argument --not-find-best-unc the code will not try to adjust the best maximum uncertainty allowed for each dataset, this optimization depends heavily on available statistics and will slow down code execution if turned on (default). Be aware that optimizing maximum allowed uncertainty impacts heavily on effiency maps shape which can be seen in efficiency per eta bin plots.