TopBox: Method

Monday, August 24, 2009

12:31

 



 



M2 and M3 using MC Matching Selction in 100/pb of ttbar single muon events.

 

Notice that the center of the M2 and M3 distributions are actually at 88 GeV and 180 GeV repectively. There could be a selection bias since not all Jets are used to reconstruct these invariant masses. The distribution above is only a subset of all semileptonic ttbar events. If we reduce the Jet Pt cut from 30 GeV to only 5 GeV, we will see a shift back to the actual W-boson and Top-quark masses as show in the plot below.

 



M2 and M3 using MC Matching Selection in 100/pb of ttbar single muon with no Jet Pt Cut.

 

Notice W-boson is still slight above the PDG define value of 80.6 GeV. There be other bias or Jet reconstruction effect causing a slightly higher "Effective W-boson Mass." However, we are experimentally constriant to use a sensible Jet Pt cut, otherwise one would be swarm by detector fluxuations since every HCAL cell passing a low read-out threshold will effectively become a Jet. From this point on we will use an effective W-mass of 90 GeV and effective Top-mass of 180 GeV.

 

From the first M2 and M3 distributions we can get an idea of the M2 and M3 cuts used to identify semileptonic  ttbar events. Using these distribution we can choose a first order cut of 70 < M2 < 110 and 150 < M3 < 210 in GeV. Using these cuts we run through all events passing the 1 muon 0 electron, 4 Jets with the leading Jet's Pt > 80 GeV.  All event which pass the M2 and M3 are consider ttbar-like and place in the TopBox as a ttbar-like sample.

TopBox: Self Validation

Thursday, August 27, 2009

13:54

Using 2/fb of data we can conduct 20 Psuedo Experiment of 100/pb.

 

First we like to check that the MET Predictions are consistent through out the different "runs."

 

In the 20 runs, each made a MET > 200 prediction and the actual values are plotted for comparison.

 

 



Pull Distribution Using old Error Calculation

 

 



Pull Distribution Using new Error Calculation

 

 



Residual Distribution.

 

 

Old Error = sqrt(scale*TopBox_HiMet) = sqrt(pred)

 

new Error = Pred* sqrt ( 1/TopBox_HiMET + 1/TopBox_LowMET + 1/All_LowMET)

 

The error all got larger using the new calculation

 

old    : new

5.3101:6.67065

6.57907:8.26456

5.2174:6.60978

6.15397:7.87936

6.02395:7.69694

5.49927:6.98513

5.60703:7.08998

6.09437:7.74008

5.32291:6.69819

6.5273:8.26419

5.52389:7.05111

5.27362:6.57211

5.98989:7.61427

6.59975:8.46124

5.58558:7.0672

6.31655:8.02265

6.14189:7.84892

5.8015:7.28212

6.56597:8.52198

6.90276:8.9971

 

All previous plots were shown with only ttbar events, however in data the selection of 1 Muon, and 4 Jets will have constributions. Some were due to mismeasurements while others are due to real processes such as W+Jets, where the W decay leptonically. In this section we will examine how much W+Jets occur in the TopBox.

Using a sample of 100/pb of W+Jets, where the W is force to decay into a muon in order to reduce the sample size. The selection cut actually reduce theW+Jets constribution to 194 events. Had-TopBox Selection further reduce this to 76 events. If Lep-TopBox is applied, the W+Jet sample is further suppressed to 45 events.

 

For now, let's use Had-TopBox, or just TopBox from now on, and look at how W+Jets affect the prediction such as MET.

 

 



Normalized MET Distribution for ttbar and W+jets. W+Jet's MET peaks slight lower than ttbar, but it seems to have a longer tail at high MET.

 



ttbar and wjets mixed together 100/pb. TopBox is scaled using 50 < MET < 100 bins and the prediction is shown below.1

 

MET > 200 GeV with W+Jets Mixed.

100/pb TBox:33.7247 Real:41

1000/pb TBox:417.213  Real:504

 

MET > 200 GeV with ttbar Only.

100/pb TBox:28.1972 Real:35

1000/pb TBox:343.055 Real:395

Friday, September 11, 2009

11:50

The single lepton select should reject most QCD contribution to the TopBox. Fake / Secondary lepton can cause a non-trivial changes to our predictions.

 

Using 1/pb(1.8M events) of QCD Pt80+ event, we like to study the effects of QCD in the sample. Scaling must be applied due to QCD's large cross-section.

 

 

 



MET Distribution for 1/pb of QCD_Pt80up which passes all selection cuts and HadTopBox.

 

Due to the limit number of events passing the single muon cut the distribution do not have enough events to be useful. We will loosen the muon cut in order to get a better distribution and scaling the distribution appropiately in order to simulate the same magnetitude of the QCD contribution.

This is done by pull a scale factor in the muon relative isolation and muon d0 cut.

 

First we should verify that this does not change MET shape of the QCD events. Looser muon cut will likely to allow secondard muons into the sample. These event will have higher MET than those with no real muon in the event.

 

 



Normalized MET Distribution for QCD using Std Muon Cuts, x5 Cut, and x10 Cut.

 

Although the statistics are low, it seems like the MET peaks higher as we allow more muons into the sample.  This would make since if we're letting in more secondary muon as oppose to random fake muons. Secondard muons produces secondary neutrino thus creating secondary MET, where fake muons should not have any real MET at all.

 

 

vector<pat::Muon> GetMuons(ChainEvent * ev, string Label, float pt_cut, float bkgFactor){

Handle<vector<pat::Muon> > handle;

handle.getByLabel(*ev,Label.c_str());

std::vector<pat::Muon> const & muons = *handle;

std::vector<pat::Muon> GoodMuons;

 

for(int i = 0; i < int(muons.size());i++){

if(muons.at(i).isTrackerMuon() || muons.at(i).isGlobalMuon()){

float d0Fix = muons.at(i).innerTrack()->d0()-0.0325839*sin(muons.at(i).innerTrack()->phi());

float TotRelIso03 =(muons.at(i).isolationR03().emEt

+muons.at(i).isolationR03().hadEt

+muons.at(i).isolationR03().sumPt)

/muons.at(i).pt();

if(        muons.at(i).pt() > pt_cut

&&        fabs(muons.at(i).eta()) < 2.1

&&  fabs(d0Fix) < 0.02*bkgFactor

&&        muons.at(i).innerTrack()->normalizedChi2() < 10

&&        muons.at(i).innerTrack()->numberOfValidHits() > 11

&&        TotRelIso03 < 0.1*bkgFactor

){

GoodMuons.push_back(muons.at(i));

}

}

}

return GoodMuons;

}

 

 

TopBox: MET Bias From Leptonic M3

Wednesday, September 02, 2009

15:18

The TopBox method can be easily expanded or contracted into more selector variable. One example is the leptonic side of ttbar decay. If we assume the MET is entirely due to the neutrino from ttbar decay, we use the W-boson's mass and the muon's  four-momentum to calculate 2 possible solution of neutrino's Pz. Then we can loop over all remaining jets in the event to find the one that reconstruct a M3 nearest to the top mass.

 

 



MET Distribution for Hadronic and Had+Lep selected events for 1/fb at 10 TeV using only ttbar sample

 

Since this sample is pure ttbar one should expect the prediction to match unless there's a selection bias in the method. If we integral the events with MET above 200 GeV, we see an unestimate of number of events with MET greater than 200 GeV.

Actual MET > 200 = 395

HadTBox MET > 200 = 343

Lep+Had TBox MET > 200 = 222

 

The Leptonic M3 cut seems to reject high MET events. A possible reason for this is because higher MET events are easier to mismeasure due to larger absolute Pt error with higher Pt Jets. High MET events also tends to have more Jets, since any mismeasurement in each Jet's energy contribution to MET.  The selection purity of the method is highly dependent of the Jet Multiplicity of the event due to combinatoric backgrounds.

 

 

 



MHT Distribution for Hadronic and Had+Lep selected events for 1/fb at 10 TeV.

 

 



Muon Pt Distribution for Hadronic and Had+Lep selected events for 1/fb at 10 TeV.

 

 

 



SumET Distribution for Hadronic and Had+Lep selected events for 1/fb at 10 TeV.

TopBox: DiLepton Effects

Friday, September 04, 2009

12:50

Thursday, September 10, 2009

11:20