FCNC fitter was originally written by Christoph Langenbruch. Then, it was modified by David Gerick to fit the needs of the B+->K*(KSpi+)mumu analysis. Later, Renata Kopecna modified it further to be slightly more versatile, not have either verbose or super verbose modes only and to compile much faster.
Table of Contents
- Compilation
- Creating necessary folders
- spdlog
- git
- Output
- Known issues
- Running
- Tuple conversion
- Parameters and constants
- Fit models
- Angular acceptance
- File structure
Compilation
The compilation is done by using cmake. The include of spdlog and all the libraries is done in CMakeLists.txt. When you install spdlog, check the path is properly included! Also, it will fail on lhcba1, run on d0new.
Compile the code with a simple
bash cmake.sh
Or just run the code using the amazing python tool for running the code, it will take care of the compilation automatically.
Creating necessary folders
There is a neat bash script for that, just run
bash makefolders.sh
spdlog
Spdlog is a fast logging library. It is made for C++ and consists only of headers. Spdlog is used in FCNCfitter now, as the previous version only used two levels of verbosity (verbose and super verbose) using basic prints. To run the current version of the FCNC fitter it is required to be installed.
The github repository can be found at https://github.com/gabime/spdlog/wiki/1.-QuickStart
For simple instalation guide and tutorials, see https://spdlog.docsforge.com/master/
After installation, check that the path to the spdlog/include and spdlog/include/spdlog folder is added to the CMakeLists.txt file. Either set it relative to the directory where the CMake is stored
set(SPDLOG_INCLUDE_DIR "${CMAKE_CURRENT_LIST_DIR}/yourRelativePath/spdlog/include/")
set(SPDLOG_INCLUDE_DIR_TWO "${CMAKE_CURRENT_LIST_DIR}/yourRelativePath/spdlog/include/spdlog/")
or set the absolute path
set(SPDLOG_INCLUDE_DIR "$/yourPath/spdlog/include/")
set(SPDLOG_INCLUDE_DIR_TWO "/yourPath/spdlog/include/spdlog/")
git
There is a python script for easier commiting and pushing to git. Check push.py for the details.
Output
There are possible 6 levels of verbosity, which follows the spdlog standard. The levels have different color codes
| level | color | description |
|---|---|---|
| 0) Trace | white | Used for development purposes, not recommended to stream this into a file because the output is then VERY verbose. |
| 1) Debug | blue | More verbose settings, prints also numbers and more detailed status of the code, shoudl be used in case of issues |
| 2) Info | green | Default settings, prints the current status of the code |
| 3) Warning | yellow | Shows not critical or small problems only |
| 4) Error | red | Prints only error messages |
| 5) Critical | full red | Prints only very serious error messages, the program should end/or crash after such message |
The plots are saved in ./plots/ folder. There are a lot of subfolders for easier organization of the plots.
Known issues
The running per year is not working flawlessly as it wasn't really ever used. The paths are sometimes wrongly configured due to this and hence the loading of files sometimes fails.
Running
For the detailed explanation of possible parameters, click here.
For quickly running everything, run the following commands
python run.py -convert -all -Run 1
python run.py -dontCompile -convert -all -Run 2
python run.py -dontCompile -convert -genMC -Run 2
python run.py -dontCompile -convert -genMC -PHSP -Run 2
python run.py -dontCompile -MC -angCorr -Run 1 -scan
python run.py -dontCompile -MC -angCorr -Run 2 -scan
python run.py -dontCompile -MC -angCorr -Run 1
python run.py -dontCompile -MC -angCorr -Run 2
python run.py -dontCompile -MC -angRes -Run 1
python run.py -dontCompile -MC -angRes -Run 2
python run.py -dontCompile -MC -Run 12 -fit -Ref -nBins 1
python run.py -dontCompile -MC -Run 12 -fit -nBins 5
python run.py -dontCompile -Run 12 -v 2 -fit -Ref -nBins 1 -genMC -Ref
python run.py -dontCompile -Run 12 -v 2 -fit -nBins 5 -genMC
python run.py -dontCompile -Data -Run 12 -fit -Ref -nBins 1 -massDim
python run.py -dontCompile -Data -Run 12 -fit -nBins 5 -massDim
python run.py -dontCompile -Data -Run 12 -fit -Ref -nBins 1 -onlyBkg -upper
python run.py -dontCompile -Data -Run 12 -fit -nBins 5 -onlyBkg -upper
python run.py -dontCompile -Data -Run 12 -v 2 -fit -Ref -nBins 1
- First, it compiles the code and prepares the tuples with Run 1.
- Then, since we already compiled the code, we use the -dontCompile option for running the next commands.
- We prepare the tuples for Run2. Then, it scans the order of polynomial used for the PHSP angular description.
- Next step is to create the angular acceptance weights using the optimal polynomial.
- Then, the signal Monte Carlo sample is fitted: first fit the reference channel, then the signal channel.
- After this step is done, it fits the generator level Mont sample in the J/psi Q2 bin and rare channel in 5 Q2 bins.
- The next step is to fit the data in the B+ and K*+ masses: first, fit the J/psi channel, then the rare channel in the 5 bins.
- Then, fit the angular background distributions (upper B+ mass sideband) using a chebyschev polynomial.
- The last step is to fit the data, J/psi channel in 4 dimensions (B+ mass, cos(thetal), cos(thetak), phi)
Tuple conversion
The functions used to convert and tne load the converted TTrees are defined in event.cc. It takes the output from the selection and converts it into a tuple with branch names that make the fitter's life easier. The correct path is saved in paths.cc, function getSelectedTuplePath(int dataset, int run, int year).
To run the conversion for all the samples, do
python run.py -convert -all -Run 1
python run.py -convert -all -Run 2
Parameters and constants
Constants
The constants are saved in constants.hh and constants.cc.
The constants in constants.cc are functions either returning a combination of global constants in constants.hh, such as
double CTL_RANGE(){
return CTL_MAX - CTL_MIN;
}
or returns values depending on number of bins, either the Q2 edges of the bins or the S and P observables.
It also contains a function returning a vector of strings with the names of the background parameters depending on the folding and the order of the used polynomial.
In constants.hh global variables are included:
- Mathematical constants (1 sigma = 0.638, pi, ... )
- Q2 ranges
- Angle ranges
- (LaTeX) names
- Plotting options
- PDG masses
- Mass windows
- Particle IDs
- Angular correction orders
- Weighting options
- Parameter values
- Quite some obsolete constants from the Kspi+ analysis
Parameters
The class definition of the parameters used in the fitter is in parameters.cc. The class consist of
- current value
- star value
- previous value
- error
- upper error
- lower error
- previous upper error
- previous lower error
The parameters used in the fitter are then defined in bu2kstarmumu_parameters.hh. The functions defined there inititialize the parameters either to the defualt or based on the given fit model. The functions can also load parameters from files, which is handy when fixing parameters to the reference channel or when generating toys. Helper functions used to retrieve the data from files are also defined there.
Options
Most of the fit options are stored in the class defined in opitons.hh. In the class init, the default values are set. As this is a legacy code, there is A LOT of unused options or options that just don't do anything.
Fit models
There are many options how to put together the model that has to be fit. Therefore, separate files with definitions of the model for given case are saved in given source files. The files that were actually used are listed in the neat table below. For the full details, consult my thesis.
| File name | Description |
|---|---|
| backgroundfit.cc | Performs the fit of the background only: taken from upper mass sideband, functions are chebyshev polynomial for angles and exponential for the B+ mass. |
| genlvlfit.cc | Loads the generator level tuple and performs the full angular fit. |
| mainfit.cc | Fits the data in 4 dimensions: B+ mass and the three angles. Be careful when running on rare data, the blinding might not be set perfectly as it never got to actually run the code |
| massfit.cc | Fits the B+ and K*+ masses. B+ mass is fitted using double sided Crystal Ball, K*+ uses the LASS parametrization |
| mcfit.cc | Performs the monte carlo fit: there is no background contribution. Fit is performed in 4D: B+ mass and the angles |
| toysfit.cc | Fits the distributions from toys. |
Angular acceptance
The angular acceptance functions are defined in angularcorr.cc. The range of the polynomials is set in the get_scan_low(bool quickTest, bool only_1D_chi2) and get_scan_high(bool quickTest, bool only_1D_chi2) functions. The actual math is done in bu2kstarmumu_pdf.cc.
File structure
Currently, the file structure is as follows. All the .cc source files also have their corresponding header with the same name.hh.
- bu2kstarmumu.cc: The main
- sources
- help.cc
- parse.cc
- paths.cc
- tests.cc
- Core: the core functions used in the FCNC fitter, where the calculations happen
- bu2kstarmumu_generator.cc
- bu2kstarmumu_loader.cc
- bu2kstarmumu_parameters.cc
- bu2kstarmumu_pdf.cc
- bu2kstarmumu_plotter.cc
- events.cc
- fitter.cc
- folder.cc (Folds the events, does not have to do anything with file folders :))
- funcs.cc
- integrals.cc
- options.cc
- toystudy.cc
- Helpers: Smaller utility functions, plotting functions
- design.cc
- helpers.cc
- Params: definition of constants, parameters used in the fit, their default values and boundaries
- constants.cc
- parameters.cc
- parameterscan.cc
- Run: Functions that initialize the fit model and then perform the fit
- angularcorr.cc
- backgroundfit.cc
- feldman_cousins.cc (Never used)
- genlvlvfit.cc
- likelihoodscan.cc (Never used)
- mainfit.cc
- massfit.cc
- mcfit.cc
- momfit.cc: Method of moments fit (Never used)
- multifit.cc (Never used)
- pulls.cc
- toysfit.cc
- Scripts: Smaller functions mostly used for plotting
- EvaluateToys.cc
- GenLvlvsMC.cc
- GetMeanError.cc
- ReferencePlots.cc
- RunningScripts.cc
- ScriptHelpers.cc