Plotting

How do I normalize histograms if my input dataset is skimmed (i.e. pre-filtered)?

You can assign to the dataset a secondary_dataset parameter, which should point to a valid dataset that can be used to do the event counting. If the secondary_dataset parameter has been defined, the plotting will automatically require the counting from this dataset instead of the one is being plotted.

I’m launching the FeaturePlot task without having previously run the MergeCategorizationStats and PreCounter tasks, so they are run automatically before running the FeaturePlot. Some error appears regarding the PUweight.

As the PUweight does not appear in basic NanoAOD, you need to add it to the RDF using its dedicated RDFModule. To do so, you should add to your FeaturePlot command --PreCounter-weights-file weights.