RooFit - Real world example for systematic errors

Dear RooFit experts,

The past three days I’ve been trying to understand RooFit and use it for the analysis I am working on.
After numerous tutorials, presentations and other materials available on the internet*. I’ve learned a bit but not enough to complete the analysis.
Therefore I am posting here to improve my understanding on the provided tools. First, let me explain what I am trying to do.

For the analysis I have two channels with each one distribution in data and 4 distributions from different MC samples: signal, bkg1, bkg2, data-driven_bkg.
I am fitting the MC templates to the data distribution in order to obtain the number of signal events. All MC samples except the data-driven background are scaled to theory expectation to begin with.
In order to get the number of signal events I’ve constructed a simple model**.

This works fine for the central measurement, but soon things become more complicated. I would like now to account for systematic uncertainties. Some of them are rate changing, either global (all MC samples) or local (only affecting a single MC sample), and others are shape changing. I could perform the measurement as shown before for each of the systematics. However, this would not include the possible correlations between them.
This lead me to have a look at HistFactory***. However, I could not find how to set my parameter of interest (POI) to the number of events from the fit. I was hoping that this setup would provide me with both a possibility to get the correlation between the systematic uncertainties as well as to combine the two channels.
Back to the RooFit code I found that the RooFit result provides a correlation histogram for the variables used to construct the model. This is what I was looking for. However, I am not sure how to use it for the calculation of the correlations between the systematic uncertainties.
Searching for ‘RooFit systematic uncertainties’ I found this procedure****. It describes how to fit a model with a systematic variation to generated events using the model for the central measurement. I don’t think this is what I need.

In general I could not find tutorials/examples that I could understand or that were close enough to what I am looking to do.
If anyone has suggestions of material I could have a look at or even a real world example of an analysis evaluating the systematic uncertainties with the correct correlation in RooFit, please let me know.

Thank you in advance.
Luke

(*)
cdsweb.cern.ch/record/1379837/fi … 11_005.pdf
indico.desy.de/getFile.py/acces … onfId=5065
indico.desy.de/getFile.py/acces … onfId=5065
indico.desy.de/getFile.py/acces … onfId=5065
cdsweb.cern.ch/record/1456844/fi … 12-016.pdf
www-ekp.physik.uni-karlsruhe.de/ … slides.pdf
www-ekp.physik.uni-karlsruhe.de/ … erview.pdf

(**)
kreczko.web.cern.ch/kreczko/Roo … n_code.txt

()
kreczko.web.cern.ch/kreczko/Roo … l1_xml.txt
kreczko.web.cern.ch/kreczko/Roo … nt_xml.txt
(
*)
www-ekp.physik.uni-karlsruhe.de/ … _tutorial/