This PR introduces a new function RateEvolution to the CombineHarvester. The function loops through the parameters in a given RooFitResults object and sets the respective parameter in the workspace to the post-fit value while leaving the other parameters at the pre-fit values. For each parameter, the yield of the processes is evaluated and stored in a map of the format
map[parname] = process_yield
where parname is the name of the parameter and process_yield is the yield of all processes in the harvester instance. The PR also adds the bindings for the python interface, where the map is accessible as a python dictionary.
This information can be used to gauge the impact of individual parameters on process yields. An example code snippet would be
bins = harvester.bin_set()
for b in bins:
bin_dict = {}
for procs in process_list:
proc_harvester = harvester.cp().bin([b]).process([procs])
if len(proc_harvester.process_set()) == 0:
print("Could not load processes '{}' for bin '{}'".format(procs, b))
continue
bin_dict[procs] = proc_harvester.RateEvolution(fit)
where process_list is the list of processes that are considered. The resulting nested dictionary can be used to create e.g. a 2D Histogram of the yield changes for the processes that are introduced by a given parameter, see this example: This functionality helped to further analyze and understand the statistical model in the HIG-19-011 and might also be interesting for other analyses.
This PR introduces a new function
RateEvolutionto the CombineHarvester. The function loops through the parameters in a given RooFitResults object and sets the respective parameter in the workspace to the post-fit value while leaving the other parameters at the pre-fit values. For each parameter, the yield of the processes is evaluated and stored in a map of the formatwhere
parnameis the name of the parameter andprocess_yieldis the yield of all processes in the harvester instance. The PR also adds the bindings for the python interface, where the map is accessible as a python dictionary.This information can be used to gauge the impact of individual parameters on process yields. An example code snippet would be
where
This functionality helped to further analyze and understand the statistical model in the HIG-19-011 and might also be interesting for other analyses.
process_listis the list of processes that are considered. The resulting nested dictionary can be used to create e.g. a 2D Histogram of the yield changes for the processes that are introduced by a given parameter, see this example: