Combination of Measurements
The combination of measurements is described in its latest version in Combination Notes pp 8 TeV, pp 2.76 TeV, from which all details about the procedures itself can be extracted. Further details about the combination of triggers within one method can be obtained from the respective analysis notes of the measurements, linked in Analysis Notes and Papers.
!!! Keep in mind that any combination of measurements is HIGHLY non-trivial and needs special care in all of its steps (uncertainties, correlations, bin-shifting, fitting, weights for combination, comparison with other measurements,...)! !!!
Correlation Coefficients of Statistical/Systematic Uncertainties
For the BLUE method, the statistical as well as systematic uncertainties of the measurements are needed as input for the combination (see Analysis Notes above).
In general, however, the statistical as well as systematic uncertainties may be correlated, for example combining PCM and PCM-EMCal but also for many other different cases (carefully think about any possibility how your measurements could be correlated in statistics+systematics -> general event cuts like SPD background rejection are the same for all measurements, although the correlation of systematics will be rather small, but still it exists).
For the case of different triggers being available for a reconstruction method, for example minimum bias and calorimeter triggers for EMCal/DCal/PHOS methods, the systematic uncertainties are found to be highly correlated in general - thus a proper consideration of correlation coefficients is already needed on the level of single methods/measurements.
An example configuration and how-to run the macro can be found here for the example of the neutral pion measurements for pp @ 8 TeV:
The macro needs an input log-file, for which you find an example configuration below.
It takes the number of measurements and then the links to the systematic files, SystematicErrorAveragedSingle*, together with the method name, the number of bins and pT ranges.
Then, the single uncertainty sources are listed from the first method with respect to the second method, in how much % of the uncertainty is uncorrelated with respect to the latter method.
The correlation macro automatically reads in all the files and calculates the correlation coefficients for all common bins of the different methods.
It outputs a *.root file with all the correlation factors as well as figures, examples shown below for the neutral pion pp@8TeV:
Not only systematic correlations can be calculated, the statistical correlation factors can also be determined using the macro and adjusting the list of parameters accordingly.
IMPORTANT NOTE
To use the correlation factors in your combination, make sure to add the cases to CombinePtPointsSpectraFullCorrMat, example:
Combination of different Measurements
The combination of different measurements is performed in the macros CombineMeson... and CombineGamma.... In the following examples are shown (many more macros are available in the AnalysisSoftware repository):
CombineMesonMeasurements8TeV.C
The input files from PCM, PCMEMC, EMC, mergedEMC and PHOS need to be fed in, generated by TaskV1/ProduceFinalResultsPatchedTriggers.C. Furthermore, the output mode for the figures ("eps") is specified as well as in which direction bin-shifting should be applied for the combined spectra ("XY"). The file containing the correlation factors for pp@8TeV needs to be hand over and some more plotting options are set with last two bools.
The macro performs the combination of all different input measurements and plots the final results, stored as .eps files as well as in .root files, see Combination Note pp 8 TeV which contains lots of the output generated by the combination macro.
CombineGammaResultsPP8TeV.C
The input files from PCM, PCMEMC, EMC need to be fed in, generated by TaskV1/CalculateGammaToPi0V4.C. Furthermore, the cocktail file is specified as well as the fits of the neutral meson spectra. The file containing the correlation factors for pp@8TeV needs to be hand over and some more plotting options are set. The number "1.28" specifies the nsigma to calculate upper limits on the direct photon spectra.
Last updated