LeptonFakes

One of the plots used to study selection criteria (and therefore fakes) is the "N-1" ("N minus one") plot. An N-1 plot is a plot of a selection criteria without that criteria applied but with all others applied. For instance, lets say we have four selection cuts for muons, one of them being isolation Pt. The isolation Pt N-1 plot has the other three selections applied to whatever sample is being studied (WW for instance) and is a histogram of the isolation Pt variable.

There is a set of cuts that are applied to all plots and all samples which are called the event cuts since they cut on general features of an event. The cuts are:

hyp_lt_p4.pt()>20 && hyp_ll_p4.pt()>10 && (hyp_lt_id*hyp_ll_id)<0 && hyp_njets == 0

Muons

Currently, N-1 plots exist for all four muon criteria (ntuple variable and current cut given in parentheses ):

• d0 is the impact parameter in the transverse plane "abs(mus_d0[hyp_ll_index])<0.25". The sign of d0 is given by the curvature of the track and therefore the charge of the particle. The abs is included so the cut is applied uniformly.

• number of tracker hits (mus_validHits[hyp_ll_index]>7)

• fit of muon track ( (mus_gfit_chi2[hyp_ll_index]/mus_gfit_ndof[hyp_ll_index])<5. ), and isolation.

• For isolation, two criteria exist. Call them iso1 and iso2. They are:

(mus_iso03_sumPt[hyp_ll_index] + mus_iso03_emEt[hyp_ll_index] + mus_iso03_hadEt[hyp_ll_index]) < 6.5

and

mus_p4[hyp_ll_index].pt()/(mus_p4[hyp_ll_index].pt() + mus_iso03_sumPt[hyp_ll_index]+mus_iso03_emEt[hyp_ll_index]+mus_iso03_hadEt[hyp_ll_index] ) >= 0.92 .

The second of the two isolation criteria is a fraction between 0 and 1. Making the histogram range from 0 to 1 created a problem for binning since root excludes the upper limit of the histogram, putting these entries in the overflow bin. To fix this, I made the histogram start at 10^-4 instead of 0, and made the upper limit 1+10^-4 still with 100 bins, 1 percent per bin. That way there is no overflow.

The plots exist for the WW ntuple (the signal) and the Wjets ntuple (the dominant background sample). For all these plots, blue is the sum of loose and tight real muons, magenta is sum of loose and tight fake muons. Real means that it satisfies ( abs(hyp_ll_id)==13 && abs(hyp_ll_mc_motherid)==24 ) (with ll -> lt for tight), and fake has abs(hyp_lt_mc_motherid)!=24 with the same id requirement.

WW N-1 plot for iso1 Wjet N-1 plot for iso1

WW N-1 plot for iso2 Wjet N-1 plot for iso2

WW N-1 plot for valid hits Wjet N-1 plot for valid hits

WW N-1 plot for global fit Wjet N-1 plot for global fit

WW N-1 plot for d0 Wjet N-1 plot for d0

Also plotted is simply the Pt (transverse momentum) of the muons with and without all of the above cuts. Both plots are in each of the files below, one for the WW sample, the other for Wjets. The left hand plot is with no cuts, the right is with all four cuts. Both have the event cuts applied. Again, blue is real (sum of loose and tight), magenta is fake (loose and tight).

A problem was discovered by looking at these plots: the fake rate is too high. The reason for this is that in these plots (and currently all other plots on this page), if the mc_motherid of a muon or an electron is not a W, it is called a fake. But a fair number of these fakes are from taus, which are actually 'real' muons--they are isolated, and since the main source of taus is Ws, they are 'real' muons. Of the 715 fakes in the WW sample, 710 are from taus, 2 are from rho mesons, and 3 are from B mesons.

As expected, the Wjet sample is more scattered. Of 5150 fakes, 4599 are taus (as verified by cutting taus out). I don't understand this, but 11 of them are from muons. How can the motherid of a muon be a muon? Possibly what happened is the muon decayed in flight to an electron and the electron was reconstructed as a muon. 1 fake is actually from a gluon. About 100 are mc_motherid = 92, which is supposed to be reserved for internal use. So I don't know what this means. 20 are rhos, 1 eta, 22 are Ks, about 260 are Ds, 1 J/psi, about 50 from Bs. 13 are from baryons. If I further require the mc_id to be a muon, thereby eliminating reconstructed muons that aren't mc muons, the number of fakes goes down to 4948, for a difference of 202.

• Pt plots with tau = fake

WW muon pt plot with and without cuts Wjet muon pt plot with and without cuts

Electrons

Here is the Pt plot for electrons. The plot on the left in each file is with no cuts, on the right, with all cuts. Both have event cuts. Red is real, green is fakes. A final note is necessary regarding the Wjets plot. When drawing histograms on top of each other, root displays the number of entries of the first one plotted in the key, ignoring the number of entries in the overlaid plot (though I'm sure I could change this if I knew how). There are way more electron fakes before cuts, but more reals after cuts, so the order of displaying is switched to make up for this. The side effect being the listed number of entries are for opposite plots. In any case, the reals before the cut have 63188 entries, and 57255 entries after. The fakes have 163948 before and 7325 after.

WW electron pt plot with and without cuts Wjet electron pt plot with and without cuts

For the N-1 Plots, there are four cuts pertinent. Two, however, have no N-1 plot becuase they cut on indicies, not continuous variables. These two are: "els_tightId[hyp_lt_index] = 1" requires that the electron pass some 'tightId' criteria (I don't know what these are though), and "els_closestMuon[hyp_lt_index] = -1" requires that there is no muon within a small (how small?) cone around the electron. This cuts down muons faking electrons. The N-1s for the remaining two cuts are:

• The isolation N-1 plots: " hyp_lt_p4.pt()/(hyp_lt_p4.pt() + els_tkIso[hyp_lt_index]) > 0.92 "

WW N-1 plot for iso Wjet N-1 plot for iso

• The d0 N-1 plots: "abs(els_d0[hyp_lt_index]) <= 0.025 ". Same comment on abs() applies as above, or moreso as this is a more important (tighter) cut for electrons.

WW N-1 plot for d0 Wjet N-1 plot for d0

• Efficiency plots: efficiency is defined as the ratio of the number which satisfy "els_tightId[hyp_lt_index] == 1" to the total number, per bin. Here efficiency is plotted as a function of pt(Gev). Again, red is real, green is fakes. Loose and tight are added before efficiency is calculated.

WW electron efficiency Wjet electron efficiency

• dR plots: the phi and eta values of a muon closest to an electron is " mus_p4[els_closestMuon[hyp_ll_index]] ", where hyp_ll(t)_index is required to be a loose (tight) electron based on cuts. The only other cut applied is " els_closestMuon[hyp_lt_index] = -1 " which requires that the closest muon is within dR = 0.1 . This is true for the first four plots. For the last plot, all the cuts are applied except, obviously, the opposite of the above cut. This is the N-1 plot. Loose and tight are added together. Red is reals, green is fakes.

WW dR between els and closest mu (dR < 0.1) Wjet dR between els and closest mu (dR < 0.1)

WW dR between els and closest mu (dR < 0.004) Wjet dR between els and closest mu (dR < 0.004)

Wjet dR between els and closest mu with other cuts (dR < 0.1)

-- WarrenAndrews - 17 May 2008