A tool to fit graphs and histograms with BAT
The idea of this code is to make it simple and easy to make Bayesian MCMC fits of Graphs and Histograms to functions. The syntax follows closely root syntax for fitting TH1D and TGraph 3 modes are currently implemented with examples in TestMacro.cxx
To compile you can use:
g++ -std=c++0x root-config --cflags --glibs macros/TestMacro.cxx src/BatGraphFitter.cxx src/BAT_GraphFit.cxx -o bin/TestMacro bat-config --libs --cflags
Or change the name of TestMacro.cxx to your macro.
To RUN the test do: ./bin/TestMacro
For the CUORE efficiencies compile and do: ./bin/CUORE_Efficiency
This will make a series of files in the output directory out_pca_pass_${ENERGY}plots.pdf - plots for the counting analysis for the peak at energy {ENERGY} out_pca_fail${ENERGY}_plots.pdf - same for events failing the cut
out_pca_pass.pdf/ out_pca_fail.pdf - plots of the fit reconstruction and posteriors on number of counts
eff.pdf - plots of the posteriors on efficiency and the graph fit eff_fit.pdf - summary plots for the Graph fit
most important
output_eff.root - root file with all 1D and 2D posteriors for further analysis eg.
KEY: TH1D h1_fit_parameter_p0;1 KEY: TH2D h2_fit_par_p1_vs_par_p0;1 KEY: TH2D h2_fit_obs_fQ_vs_par_p0;1 KEY: TH1D h1_fit_parameter_p1;1 KEY: TH2D h2_fit_obs_fQ_vs_par_p1;1 KEY: TH1D h1_fit_observable_fQ;1
// ** BELOW IS AN EXPLANTION OF THE BAT CODE // ---------------------------------------------------------------------------------------------------------
- Fit a graph assuming gaussian errors
Syntax is something like:
BatGraphFitter *fitter = new BatGraphFitter(g,f); fitter->SetPrecison(3); // sets the precion with 1 =kLow, 2 = kMedium, 3=kHigh, 4 = kVeryHigh (generally 2 or 3 are good choices)
fitter->Fit("","",0,3000);
ie the last Fit command has the same syntax as ROOT
sometimes you have constraints eg the data are effieciencies or resoltution, to restrict to the physical region use fitter->SetGraphMaxMin(double max,double min); To limit the data to be within this range (multiply by this uniform distribution)
- Fit a graph assuming a binomial error (for some efficiencies)
Constuct the fitter like
BatGraphFitter(TGraph *g,TGraph *&gTrial,TGraph *&gSuccess,TF1 *&f);
WHere gTrial is the number of throws, gSuccess the number of catch and g the ratio (efficiency)
- Fit a histogram assuming Poisson errors
Construct with BatGraphFitter *fitter = new BatGraphFitter(fHist,func); fitter->SetPrecison(3); fitter->Fit();
Currently accesing the outputs is in the "standard way" for BAT, I will try to add more methods soon
eg. // to print the marginalised posterior or write the markov chain do this fitter->fModel->PrintAllMarginalized("out_plots.pdf"); fitter->fModel->WriteMarkovChain("output_mcmc.root", "RECREATE");
// a plot of the best fit will automatically be produced and can be saves