Hello! Is there any option in FCCAnalyses or plan to include statistical/systematic uncertainties in the histogram outputs from the plots step? e.g.
fccanalysis plots analysis_plots.py
I haven’t seen any output plots including this, and have not encountered an option to do so yet. I suppose syst. uncertainties would require additional ntuples w.r.t. the nominal ones, but statistical uncertainties should be implementable right?
Also, is this FCCAnalysis category the appropriate channel for questions about the FCCAnalyses repository?
Thanks for reaching out. Yes, you are at the right place!
There is no such option that I know of at the moment but yes it could be added.
Just to be sure, you refer to having a greyed hashed band on top of the MC stack to display statistical uncertainties on the plot itself, not to produce the series of TH1s with single bin fluctuated up/down to treat the bin-by-bin uncertainty in the final fit? The latter functionality should probably be provided by the fitting tool used.
Would something like the plot attached meet your need?
Adding other systematic uncertainties on top of the MC statistics one, even though they come from a different set of TH1s, could also be implemented.
Thanks for the reply - what you have attached from CMS is exactly what I had in mind
not to produce the series of TH1s with single bin fluctuated up/down to treat the bin-by-bin uncertainty in the final fit? The latter functionality should probably be provided by the fitting tool used.
Indeed - we would need that functionality to be in whatever fitting tool used. I just think having the stat uncertainties in the plots is important to visualize and understand how much stat uncertainty we expect in the results of a given phase space.
We can certainly implement the option to display the MC stat uncertainties, I’ll have a look into that and I’ll let you know. Though for FCC at the Higgs pole, we expect the MC stat uncertainty should certainly be negligible w.r.t. the data statistics, so if we lack MC statistics we just need to produce more samples.
We can certainly implement the option to display the MC stat uncertainties, I’ll have a look into that and I’ll let you know.
Thanks a lot!
Though for FCC at the Higgs pole, we expect the MC stat uncertainty should certainly be negligible w.r.t. the data statistics, so if we lack MC statistics we just need to produce more samples.
What do you mean by “data statistics”? Are you really referring to future data to be taken at FCC? My thinking is that even if we have a fair amount of MC stats to start, it would be helpful to visualize the stat uncertainty to see when it becomes more significant e.g. after tighter kinematic / MVA selections.
Indeed, I meant the data statistics of the actual data to be taken at the FCC. But you are right, it is useful to check how the MC stats behave in certain phase spaces which can give us an idea.
The option to display the statistical uncertainty is implemented now, see an example here:
