using CMSSW_6_1_1 on lxplus5
cmsenv
make

1. Prepare the training and testing files:
mkdir files
<copy flat tree with desired signal samples to files/tree_Signal.root>
edit src/RegVars.C to (un)comment the desired inTree name (choose between MSSM or ggHH). remake.
./RegVars <mass>
mass is the resonant mass. this value doesn't matter for the ggHH sample, so use whatever.

2. Create the weight file:
root -l -b -q "Regression/TMVARegression.C(\"SM\")" #for the nonresonant sample
root -l -b -q "Regression/TMVARegression.C(\"resonant\")" #for the resonant sample
if step 1 was run correctly, Regression/TMVARegression.C should not need to be changed.

3. Add the variables
./AddRegVars MSSM_H_m260_8TeV 1
./AddRegVars MSSM_H_m300_8TeV 1
./AddRegVars MSSM_H_m350_8TeV 1
./AddRegVars Radion_m270_8TeV -1
./AddRegVars Radion_m300_8TeV -1
./AddRegVars Radion_m350_8TeV -1
./AddRegVars Radion_m400_8TeV -1
./AddRegVars ggHH_8TeV 1
the first argument is the tree name.
the second designates which subset of events to analyze. -1 for all, 0 for event only, 1 for odd only.
step 2 trains on even events, so only odd events should be considered. Radion is not used for training.
AddRegVars keeps events in the 2tag category.

4. Create plots and evaluate improvement
src/RegVal.C can be edited to change some options, but in general it shouldn't need to be changed.
./RegVal MSSM_H_m260_8TeV 260
./RegVal MSSM_H_m300_8TeV 300 #mv to avoid overwriting
mv plots/resImp_voigt_M-300.png plots/resImp_voigt_M-300_MSSM.png 
mv plots/resImp_gauss_M-300.png plots/resImp_gauss_M-300_MSSM.png 
mv plots/resImp_gauscb_M-300.png plots/resImp_gauscb_M-300_MSSM.png 
./RegVal MSSM_H_m350_8TeV 350 #mv to avoid overwriting
mv plots/resImp_voigt_M-350.png plots/resImp_voigt_M-350_MSSM.png 
mv plots/resImp_gauss_M-350.png plots/resImp_gauss_M-350_MSSM.png 
mv plots/resImp_gauscb_M-350.png plots/resImp_gauscb_M-350_MSSM.png 
./RegVal Radion_m270_8TeV 270
./RegVal Radion_m300_8TeV 300
./RegVal Radion_m350_8TeV 350
./RegVal Radion_m400_8TeV 400
./RegVal ggHH_8TeV -1