def ana(files,returnplots=False):
#%%################
# Plots and Setup #
###################
## Make a dictionary of histogram objects
plots = {
"RjetpT": Hist(100,(0,100) ,'RECO matched jet pT','Events','recplots/RjetpT'),
"Rjeteta": Hist(66 ,(-3.3,3.3) ,'RECO matched jet eta','Events','recplots/Rjeteta'),
"RjetCSVV2":Hist([0,0.1241,0.4184,0.7527,1],None,'RECO matched jet btagCSVV2 score','events','recplots/RjetCSVV2'),
#"RjetDeepB":Hist([0,0.0494,0.2770,0.7264,1],None,'RECO matched jet btagDeepB score','events','recplots/RjetDeepB'),
"RjetDeepFB":Hist([0,0.0494,0.2770,0.7264,1],None,'RECO matched jet btagDeepFlavB score','events','recplots/RjetDeepFB'),
"RA1pT": Hist(80 ,(0,160) ,'pT of RECO A1 objects constructed from matched jets','Events','recplots/RA1pT'),
"RA2pT": Hist(80 ,(0,160) ,'pT of RECO A2 objects constructed from matched jets','Events','recplots/RA2pT'),
"RA1mass": Hist(40 ,(0,80) ,'reconstructed mass of A1 objects from matched jets','Events','recplots/RA1mass'),
"RA2mass": Hist(40 ,(0,80) ,'reconstructed mass of A2 objects from matched jets','Events','recplots/RA2mass'),
"RA1dR": Hist(50 ,(0,5) ,'dR between jet children of reconstructed A1 object','Events','recplots/RA1dR'),
"RA2dR": Hist(50 ,(0,5) ,'dR between jet children of reconstructed A2 object','Events','recplots/RA2dR'),
"RA1deta": Hist(33 ,(0,3.3) ,'|deta| between jet children of reconstructed A1 object','Events','recplots/RA1deta'),
"RA2deta": Hist(33 ,(0,3.3) ,'|deta| between jet children of reconstructed A2 object','Events','recplots/RA2deta'),
"RA1dphi": Hist(33 ,(0,3.3) ,'|dphi| between jet children of reconstructed A1 object','Events','recplots/RA1dphi'),
"RA2dphi": Hist(33 ,(0,3.3) ,'|dphi| between jet children of reconstructed A2 object','Events','recplots/RA2dphi'),
"RHmass": Hist(80 ,(0,160) ,'reconstructed mass of Higgs object from reconstructed As','Events','recplots/RHmass'),
"RHpT": Hist(100,(0,200) ,'pT of reconstructed higgs object from reconstructed As','Events','recplots/RHpT'),
"RHdR": Hist(50 ,(0,5) ,'dR between A children of reconstructed higgs object','Events','recplots/RHdR'),
"RHdeta": Hist(33 ,(0,3.3) ,'|deta| between A children of reconstructed higgs object','Events','recplots/RHdeta'),
"RHdphi": Hist(33 ,(0,3.3) ,'|dphi| between A children of reconstructed higgs object','Events','recplots/RHdphi'),
##
"RalljetpT": Hist(100,(0,100),'All RECO jet pT','Events','recplots/RalljetpT'),
"npassed": Hist(1 ,(0.5,1.5) ,'','Number of events that passed cuts', 'recplots/npassed')
}
for plot in plots:
plots[plot].title = files[0]
## Create an internal figure for pyplot to write to
plt.figure(1)
## Loop over input files
for fnum in range(len(files)):
#####################
# Loading Variables #
#####################
print('Opening '+files[fnum])
## Open our file and grab the events tree
f = uproot.open(files[fnum])#'nobias.root')
events = f.get('Events')
jets = PhysObj('jets')
jets.eta= pd.DataFrame(events.array('Jet_eta')).rename(columns=inc)
jets.phi= pd.DataFrame(events.array('Jet_phi')).rename(columns=inc)
jets.pt = pd.DataFrame(events.array('Jet_pt')).rename(columns=inc)
jets.mass=pd.DataFrame(events.array('Jet_mass')).rename(columns=inc)
jets.CSVV2 = pd.DataFrame(events.array('Jet_btagCSVV2')).rename(columns=inc)
jets.DeepB = pd.DataFrame(events.array('Jet_btagDeepB')).rename(columns=inc)
jets.DeepFB= pd.DataFrame(events.array('Jet_btagDeepFlavB')).rename(columns=inc)
print('Processing ' + str(len(jets.eta)) + ' events')
## Figure out how many bs and jets there are
njet= jets.eta.shape[1]
if njet > 10:
njet = 10
ev = Event(jets)
jets.cut(jets.pt>15)
jets.cut(abs(jets.eta)<2.4)
jets.cut(jets.DeepB > 0.4184 )
ev.sync()
##############################
# Processing and Calculation #
##############################
## Create our dR dataframe by populating its first column and naming it accordingly
jjdr2 = pd.DataFrame(np.power(jets.eta[1]-jets.eta[2],2) + np.power(jets.phi[1]-jets.phi[2],2)).rename(columns={0:'Jet 1 x Jet 2'})
jjmass = pd.DataFrame(jets.mass[1] + jets.mass[2]).rename(columns={0:'Jet 1 x Jet 2'})
## Loop over jet x b combinations
jjstr = []
for j in range(1,njet+1):
for i in range(j+1,njet+1):
## Make our column name
jjstr.append("Jet "+str(j)+" x Jet "+str(i))
#if (i+j == 3):
# continue
## Compute and store the dr of the given b and jet for every event at once
jjdr2[jjstr[-1]] = pd.DataFrame(np.power(jets.eta[j]-jets.eta[i],2) + np.power(jets.phi[j]-jets.phi[i],2))
jjmass[jjstr[-1]] = pd.DataFrame(jets.mass[j] + jets.mass[i])
#if (j==i):
# jjdr2[jjstr[-1]] = jjdr2[jjstr[-1]] * np.nan
# jjmass[jjstr[-1]] = jjmass[jjstr[-1]] * np.nan
print('jjs done')
j4mass = pd.DataFrame(jets.mass[1]+jets.mass[2]+jets.mass[3]+jets.mass[4]).rename(columns={0:"J1 x J2 J3 J4"})
j4str = []
for a in range(1,njet+1):
for b in range(a+1,njet+1):
for c in range(b+1,njet+1):
for d in range(c+1,njet+1):
j4str.append("J"+str(a)+" J"+str(b)+" J"+str(c)+" J"+str(d))
#if (a+b+c+d == 10):
# continue
j4mass[j4str[-1]] = pd.DataFrame(jets.mass[a]+jets.mass[b]+jets.mass[c]+jets.mass[d])
#if (a==b or a==c or a==d or b==c or b==d or c==d):
# j4mass[j4str[-1]] = j4mass[j4str[-1]] * np.nan
print('j4s done')
## Create a copy array to collapse in jets into
drlist = []
mmlist = []
m4list = []
for j in range(njet):
drlist.append(np.sqrt(jjdr2.filter(like='Jet '+str(j+1))))
mmlist.append(jjmass.filter(like='Jet '+str(j+1)))
m4list.append(j4mass.filter(like='J'+str(j+1)))
#jlist[j] = jlist[j][jlist[j].rank(axis=1,method='first') == 1]
#drlist[j] = jlist[j].rename(columns=lambda x:int(x[4:6]))
#mmlist[j] = mmlist[j].rename(columns=lambda x:int(x[4:6]))
print('jlist done')
## Cut our events to only resolved 4jet events with dR<0.4
djets = mmlist[0][(mmlist[0]>25) & (mmlist[0]<65)]
qjets = m4list[0][(m4list[0]>90) & (m4list[0]<150)]
for i in range(1,njet):
djets = djets.combine_first(mmlist[i][(mmlist[i]>25) & (mmlist[i]<65)])
qjets = qjets.combine_first(m4list[i][(m4list[i]>90) & (m4list[i]<150)])
djets = djets / djets
qjets = qjets / qjets
djets = djets.sum(axis=1)
qjets = qjets.sum(axis=1)
djets = djets[djets>=2].dropna()
qjets = qjets[qjets>=1].dropna()
jets.trimTo(djets)
jets.trimTo(qjets)
ev.sync()
print('trimming done')
#############################
# Constructing RECO objects #
#############################
for prop in ['bpt','beta','bphi','bmass']:
jets[prop] = pd.DataFrame()
for i in range(1,5):
jets[prop][i] = jets[prop[1:]][jets.mass.rank(axis=1,method='first',ascending=False) == i].max(axis=1)
#jets.bdr = pd.DataFrame()
#for i in range(nb):
# jets.bdr[i+1] = blist[i][blist[i]>0].max(axis=1)
#ev.sync()
bvec = []
for i in range(1,5):
bvec.append(TLorentzVectorArray.from_ptetaphim(jets.bpt[i],jets.beta[i],jets.bphi[i],jets.bmass[i]))
avec = []
for i in range(0,4,2):
avec.append(bvec[i]+bvec[i+1])
for prop in ['apt','aeta','aphi','amass']:
jets[prop] = pd.DataFrame()
for i in range(2):
jets.apt[i+1] = avec[i].pt
jets.aeta[i+1] = avec[i].eta
jets.aphi[i+1] = avec[i].phi
jets.amass[i+1]= avec[i].mass
for prop in ['apt','aeta','aphi','amass']:
jets[prop].index = jets.pt.index
hvec = [avec[0]+avec[1]]
for prop in ['hpt','heta','hphi','hmass']:
jets[prop] = pd.DataFrame()
jets.hpt[1] = hvec[0].pt
jets.heta[1] = hvec[0].eta
jets.hphi[1] = hvec[0].phi
jets.hmass[1]= hvec[0].mass
for prop in ['hpt','heta','hphi','hmass']:
jets[prop].index = jets.eta.index
################
# Filling Data #
################
plots['RalljetpT'].dfill(jets.pt)
#plots['bjdR'].dfill(jets.bdr)
plots['RjetpT'].dfill(jets.bpt)
plots['Rjeteta'].dfill(jets.beta)
for i in range(1,3):
plots['RA'+str(i)+'pT' ].fill(jets.apt[i])
plots['RA'+str(i)+'mass'].fill(jets.amass[i])
#lots['RA'+str(i)+'deta'].fill(abs(jets.beta[2*i]-jets.beta[(2*i)-1]))
plots['RA'+str(i)+'dR' ].fill(np.sqrt(np.power(jets.beta[2*i]-jets.beta[(2*i)-1],2)+np.power(jets.bphi[2*i]-jets.bphi[(2*i)-1],2)))
plots['RA'+str(i)+'deta'].fill(abs(jets.beta[2*i]-jets.beta[(2*i)-1]))
plots['RA'+str(i)+'dphi'].fill(abs(jets.bphi[2*i]-jets.bphi[(2*i)-1]))
plots['RHpT' ].fill(jets.hpt[1])
plots['RHmass'].fill(jets.hmass[1])
plots['RHdR' ].fill(np.sqrt(np.power(jets.aeta[2]-jets.aeta[1],2)+np.power(jets.aphi[2]-jets.aphi[1],2)))
plots['RHdeta'].fill(abs(jets.aeta[2]-jets.aeta[1]))
plots['RHdphi'].fill(abs(jets.aphi[2]-jets.aphi[1]))
plots['npassed'].fill(jets.hpt[1]/jets.hpt[1])
plots['RjetCSVV2'].dfill(jets.CSVV2)
plots['RjetDeepFB'].dfill(jets.DeepFB)
############
# Plotting #
############
plt.clf()
#plots.pop('bjdR').plot(logv=True)
for p in plots:
plots[p].plot()
#%%
if returnplots==True:
return plots
else:
sys.exit()
def trig(files):
## Create a dictionary of histogram objects
plots = {
'cutflow': Hist(4,(-0.5,4.5),'Total / Passed 4jet pT and |eta| cut/ passed DeepB > 0.4184','Events','recplots/datacutflow'),
"RjetCSVV2":Hist([0,0.1241,0.4184,0.7527,1],None,'RECO matched jet btagCSVV2 score','events','recplots/dataCSVV2'),
#"RjetDeepB":Hist([0,0.0494,0.2770,0.7264,1],None,'RECO matched jet btagDeepB score','events','recplots/RjetDeepB'),
"RjetDeepFB":Hist([0,0.0494,0.2770,0.7264,1],None,'RECO matched jet btagDeepFlavB score','events','recplots/dataDeepFB'),
}
for plot in plots:
plots[plot].title = files[0]
## Create an internal figure for pyplot to write to
plt.figure(1)
## Loop over all input files
for fnum in range(len(files)):
print('Opening '+files[fnum])
## Open the file and retrieve our key branches
f = uproot.open(files[fnum])
events = f.get('Events')
jets = PhysObj('jets')
jets.eta= pd.DataFrame(events.array('Jet_eta')).rename(columns=inc)
jets.phi= pd.DataFrame(events.array('Jet_phi')).rename(columns=inc)
jets.pt = pd.DataFrame(events.array('Jet_pt')).rename(columns=inc)
jets.mass=pd.DataFrame(events.array('Jet_mass')).rename(columns=inc)
jets.CSVV2 = pd.DataFrame(events.array('Jet_btagCSVV2')).rename(columns=inc)
jets.DeepB = pd.DataFrame(events.array('Jet_btagDeepB')).rename(columns=inc)
jets.DeepFB= pd.DataFrame(events.array('Jet_btagDeepFlavB')).rename(columns=inc)
print('Processing ' + str(len(jets.eta)) + ' events')
## Figure out how many bs and jets there are
njet= jets.eta.shape[1]
## Fill 0 bin of cut flow plots
plots['cutflow'].fill(jets.pt.max(axis=1)*0)
ev = Event(jets)
jets.cut(abs(jets.eta)<2.4)
jets.cut(jets.pt>15)
resjets = (jets.pt/jets.pt).sum(axis=1)
resjets = resjets[resjets>=4]
plots['cutflow'].fill(resjets/resjets)
plots['RjetCSVV2'].dfill(jets.CSVV2)
plots['RjetDeepFB'].dfill(jets.DeepFB)
jets.cut(jets.DeepB > 0.4184 )
tagjets = (jets.DeepB/jets.DeepB).sum(axis=1)
tagjets = tagjets[tagjets>=2]
plots['cutflow'].fill(tagjets*2/tagjets)
ev.sync()
for pl in plots:
plots[pl].plot()
print(plots['cutflow'][0],plots['cutflow'][1])
sys.exit()
def compare(conf, option, stage):
#%%
print(f"Analysing {conf} with {option} cuts")
with open(conf) as f:
confd = json.load(f)
islhe = confd['islhe']
isdata = confd['isdata']
files = confd['files']
if type(files) != list:
files = [files]
fweights = confd['weight']
if type(fweights) != list:
fweights = [fweights]
name = confd['name']
if stage == "A":
numplot = {
'pt':
Hist(27, (150, 1500),
'pT of AK8 jet passing cuts + triggers / passing cuts',
f"{option} Ratio",
f"Effplots/{name}_pTEfficiencyPlot_{option}_A"),
}
elif stage == "B":
numplot = {
'pt':
Hist(50, (0, 1000),
'pT of AK8 jet passing cuts + triggers / passing cuts',
f"{option} Ratio",
f"Effplots/{name}_pTEfficiencyPlot_{option}_B"),
'msoft':
Hist(
11, (90, 200),
'softdrop mass of AK8 jet above 400GeV passing cuts+triggers / cuts',
f"{option} Ratio",
f"Effplots/{name}_msoftEfficiencyPlot_{option}_B"),
'DDBvL':
Hist(
20, (0.8, 1.0),
'DDBvL of AK8 jet above 400GeV passing cuts+triggers / passing cuts',
f"{option} Ratio",
f"Effplots/{name}_ddbvlEfficiencyPlot_{option}_B"),
}
elif stage == "C":
numplot = {
'pt':
Hist(27, (150, 1500),
'pT of AK8 jet passing cuts + triggers / passing cuts',
f"{option} Ratio",
f"Effplots/{name}_pTEfficiencyPlot_{option}_C"),
's2pt':
Hist(
61, (30, 1050),
'pT of 2nd highest pT slimjet passing cuts+triggers / passing cuts',
f"{option} Ratio",
f"Effplots/{name}_s2pTEfficiencyPlot_{option}_C"),
'lowb':
Hist(
20, (0, 1.0),
'Lowest deepB of two slimjets passing cuts+triggers / passing cuts',
f"{option} Ratio",
f"Effplots/{name}_lowbEfficiencyPlot_{option}_C"),
}
elif stage == "D":
numplot = {
'pt':
Hist(27, (150, 1500),
'pT of AK8 jet passing cuts + triggers / passing cuts',
f"{option} Ratio",
f"Effplots/{name}_pTEfficiencyPlot_{option}_D"),
}
elif len(stage) > 1:
numplot = {
'pt':
Hist(5, (400, 650), 'pT of AK8 jet passing cuts+triggers / cuts',
f"{option} Ratio",
f"Seffplots/{name}_{stage}_{option}_ptScalePlot"),
'msoft':
Hist(6, (80, 200), 'msoft of AK8 jet passing cuts+triggers / cuts',
f"{option} Ratio",
f"Seffplots/{name}_{stage}_{option}_msoftScalePlot"),
'DDBvL':
Hist(4, (.80, 1), 'DDBvL of AK8 jet passing cuts+triggers / cuts',
f"{option} Ratio",
f"Seffplots/{name}_{stage}_{option}_ddbvlScalePlot"),
}
# numplot['pt'][1][-1] = np.inf
# numplot['msoft'][1][0] = 90
if "C" in stage:
numplot.update({
's2pt':
Hist(
4, (140, 220),
'pT of 2nd highest pT slimjet passing cuts+triggers / cuts',
f"{option} Ratio",
f"Seffplots/{name}_{stage}_{option}_s2ptScalePlot"),
's2deepb':
Hist(
[.4184, .5856, .7527, 0.8764, 1], None,
'DeepB of 2nd highest DeepB slimjet passing cuts+triggers / cuts',
f"{option} Ratio",
f"Seffplots/{name}_{stage}_{option}_s2deepbScalePlot"),
})
if stage == "CX":
numplot.update({
'pt':
Hist(3, (250, 400),
'pT of AK8 jet passing cuts+triggers / cuts',
f"{option} Ratio",
f"Seffplots/{name}_{stage}_{option}_ptScalePlot"),
})
# for p in numplot:
# numplot[p][1][0] = 0
for p in numplot:
numplot[p].title = f"{name} {stage}"
numplot[p].ylim = (0, 1)
denplot = cp.deepcopy(numplot)
elecvars = ['pt', 'eta', 'mvaFall17V2Iso_WP90']
muvars = [
'pt', 'eta', 'mediumPromptId', 'miniIsoId', 'softId', 'dxy', 'dxyErr',
'ip3d'
]
l1vars = [
'SingleJet180', 'Mu7_EG23er2p5', 'Mu7_LooseIsoEG20er2p5',
'Mu20_EG10er2p5', 'SingleMu22', 'SingleMu25',
'DoubleJet112er2p3_dEta_Max1p6', 'DoubleJet150er2p5'
]
hltvars = [
'AK8PFJet500', 'Mu8_TrkIsoVVL_Ele23_CaloIdL_TrackIdL_IsoVL_DZ',
'Mu23_TrkIsoVVL_Ele12_CaloIdL_TrackIdL_IsoVL', 'Mu27_Ele37_CaloIdL_MW',
'Mu37_Ele27_CaloIdL_MW',
'AK8PFJet330_TrimMass30_PFAK8BoostedDoubleB_np4',
'DoublePFJets116MaxDeta1p6_DoubleCaloBTagDeepCSV_p71'
]
slimvars = ['pt', 'eta', 'phi', 'btagDeepB', 'puId']
print("Collecting event information")
events, jets, elecs, mus, l1s, hlts, sjets = [], [], [], [], [], [], []
for i in range(len(files)):
events.append(uproot.open(files[i]).get('Events'))
jets.append(loadjets(PhysObj(f"Jets{i}"), events[i], islhe))
jets[i].extweight = jets[i].extweight * fweights[i]
elecs.append(
PhysObj(f"Electron{i}", files[i], *elecvars, varname='Electron'))
mus.append(PhysObj(f"Muon{i}", files[i], *muvars, varname='Muon'))
mus[i].eta = abs(mus[i].eta)
mus[i].ip = abs(mus[i].dxy / mus[i].dxyErr)
l1s.append(PhysObj(f"L1{i}", files[i], *l1vars, varname='L1'))
hlts.append(PhysObj(f"HLT{i}", files[i], *hltvars, varname='HLT'))
sjets.append(PhysObj(f"Slimjet{i}", files[i], *slimvars,
varname='Jet'))
evs = []
for i in range(len(files)):
evs.append(Event(jets[i], l1s[i], hlts[i], elecs[i], mus[i]))
if "C" in stage:
evs[i].register(sjets[i])
for jet in jets:
jet.cut(jet.pt > 170)
jet.cut(abs(jet.eta) < 2.4)
jet.cut(jet.DDBvL > 0.8)
jet.cut(jet.DeepB > 0.4184)
jet.cut(jet.msoft > 90)
jet.cut(jet.mass > 90)
jet.cut(jet.msoft < 200)
jet.cut(jet.npvsG >= 1)
if "AB" in stage:
jet.cut(jet.pt >= 400)
elif stage == "CX":
jet.cut(jet.pt >= 250)
jet.cut(jet.pt < 400)
if option == 'MuonEG':
for elec in elecs:
elec.cut(elec.pt > 15)
elec.cut(abs(elec.eta) < 2.5)
elec.cut(elec.mvaFall17V2Iso_WP90 > 0.9)
for mu in mus:
if option == 'MuonEG':
mu.cut(mu.pt > 10)
mu.cut(abs(mu.eta) < 2.4)
mu.cut(mu.mediumPromptId > 0.9)
mu.cut(mu.miniIsoId >= 2)
elif option == 'Parked':
mu.cut(mu.softId > 0.9)
mu.cut(abs(mu.eta) < 2.4)
mu.cut(mu.pt > 7)
mu.cut(mu.ip > 2)
mu.cut(mu.ip3d < 0.5)
#else: raise(NameError("Dataset name does not match expected"))
for ev in evs:
ev.sync()
if option == 'MuonEG':
for i in range(len(files)):
l1s[i].cut(
np.logical_or.reduce(
(np.logical_and(
np.logical_or(l1s[i].Mu7_EG23er2p5,
l1s[i].Mu7_LooseIsoEG20er2p5),
hlts[i].Mu8_TrkIsoVVL_Ele23_CaloIdL_TrackIdL_IsoVL_DZ),
np.logical_and(
np.logical_or(l1s[i].Mu20_EG10er2p5,
l1s[i].SingleMu22),
hlts[i].Mu23_TrkIsoVVL_Ele12_CaloIdL_TrackIdL_IsoVL),
np.logical_and(l1s[i].SingleMu25,
hlts[i].Mu27_Ele37_CaloIdL_MW),
np.logical_and(l1s[i].SingleMu25,
hlts[i].Mu37_Ele27_CaloIdL_MW))))
## Makes a frame whose elements have the highest pt of the muon or electron in that position
passf = elecs[i].pt.combine(mus[i].pt, np.maximum, fill_value=0)
## Drops pt < 25
passf = passf[passf > 25]
## Drops empty rows
passf = passf.dropna(how='all')
## The remaining events must have had an electron or muon with pt > 25 - the rest are removed
elecs[i].trimTo(passf)
for ev in evs:
ev.sync()
if "C" in stage or stage == "AB":
for i in range(len(files)):
print(f"Processing file {i} slimjets")
jets[i].cut(
jets[i].pt.rank(axis=1, method='first', ascending=False) == 1)
sjets[i].cut(abs(sjets[i].eta) < 2.4)
sjets[i].cut(sjets[i].pt > 30)
sjets[i].cut(sjets[i].puId >= 1)
sjets[i].cut(sjets[i].pt > 140)
sjets[i].cut(sjets[i].btagDeepB > 0.4184)
evs[i].sync()
## This entire block is designed to remove any events whose defined a and b jets
## have a dR > 0.8 to the highest pT passing jet
print("Computing dR")
if sjets[i].pt.shape[0] < 1 or jets[i].pt.shape[0] < 1:
continue
sjjdr = computedR(jets[i], sjets[i], ['Fatjet', 'slimjet'])
jlist = []
print("Assembling slim frame")
for j in range(jets[i].pt.shape[1]):
jlist.append(sjjdr.filter(like=f"Fatjet {j+1}"))
jlist[j] = jlist[j].rename(columns=lambda x: int(x[-2:]))
jlist[0][jlist[0] == 0] = jlist[0] + 0.001
sjframe = jlist[0][jlist[0] < 0.8].fillna(0)
for j in range(1, jets[i].pt.shape[1]):
jlist[j][jlist[j] == 0] = jlist[j] + 0.001
sjframe = sjframe + jlist[j][jlist[j] < 0.8].fillna(0)
sjets[i].cut(sjframe != 0)
## Trims the collection of slimjets < dR 0.8 to only events with a 2nd passing jet
sjets[i].trimto(sjets[i].cut(sjets[i].pt.rank(
axis=1, method='first', ascending=False) == 2,
split=True).pt)
if stage == "AB":
for elem in sjets[i]:
evs[i].frame = evs[i].frame.loc[
evs[i].frame.index.difference(sjets[i][elem].index)]
for ev in evs:
ev.sync()
print("Assembling finished frame")
framepieces = []
for i in range(len(files)):
tempframe = pd.DataFrame()
for prop in jets[i]:
tempframe[prop] = jets[i][prop][jets[i]['pt'].rank(
axis=1, method='first', ascending=False) == 1].max(axis=1)
if option == "MuonEG" or option == "Parked":
for prop in mus[i]:
tempframe[f"m{prop}"] = mus[i][prop][mus[i]['pt'].rank(
axis=1, method='first', ascending=False) == 1].max(axis=1)
if "C" in stage:
for prop in sjets[i]:
tempframe[f"s1{prop}"] = sjets[i][prop][sjets[i]['pt'].rank(
axis=1, method='first', ascending=False) == 1].max(axis=1)
tempframe[f"s2{prop}"] = sjets[i][prop][sjets[i]['pt'].rank(
axis=1, method='first', ascending=False) == 2].max(axis=1)
tempframe['trigA'] = np.logical_and(l1s[i].SingleJet180[1],
hlts[i].AK8PFJet500[1])
tempframe['trigB'] = np.logical_and(
l1s[i].SingleJet180[1],
hlts[i].AK8PFJet330_TrimMass30_PFAK8BoostedDoubleB_np4[1])
tempframe['trigC'] = np.logical_and(
np.logical_or(l1s[i].DoubleJet112er2p3_dEta_Max1p6[1],
l1s[i].DoubleJet150er2p5[1]),
hlts[i].DoublePFJets116MaxDeta1p6_DoubleCaloBTagDeepCSV_p71[1])
tempframe['trigAB'] = np.logical_or(tempframe['trigA'],
tempframe['trigB'])
tempframe['trigABC'] = np.logical_or(tempframe['trigAB'],
tempframe['trigC'])
# tempframe['L1_SingleJet180'] = l1s[i].SingleJet180[1]
# tempframe['HLT_AK8PFJet500'] = hlts[i].AK8PFJet500[1]
# tempframe['HLT_AK8PFJet330_TrimMass30_PFAK8BoostedDoubleB_np4'] = hlts[i].AK8PFJet330_TrimMass30_PFAK8BoostedDoubleB_np4[1]
# tempframe['L1_DoubleJet112er2p3_dEta_Max1p6'] = l1s[i].DoubleJet112er2p3_dEta_Max1p6[1]
# tempframe['L1_DoubleJet150er2p5'] = l1s[i].DoubleJet150er2p5[1]
# tempframe['HLT_DoublePFJets116MaxDeta1p6_DoubleCaloBTagDeepCSV_p71'] = hlts[i].DoublePFJets116MaxDeta1p6_DoubleCaloBTagDeepCSV_p71[1]
framepieces.append(tempframe)
mergedframe = pd.concat(framepieces, ignore_index=True)
mergedframe = mergedframe.dropna()
if conf == "GGH_HPT.json":
sigweight = (3.9 - 0.4 * np.log2(mergedframe['pt']))
sigweight[sigweight < 0.1] = 0.1
mergedframe['extweight'] = mergedframe['extweight'] * sigweight
if option == 'Parked' and not isdata:
mergedframe['extweight'] = lumipucalc(mergedframe)
# pickle.dump(evs,open('effevents.p','wb'))
# pickle.dump(mergedframe,open('effframe.p','wb'))
print("Producing histograms")
effplot = {}
if stage == "A":
denplot['pt'].fill(mergedframe['pt'], mergedframe['extweight'])
numplot['pt'].fill(mergedframe['pt'][mergedframe['trigA'] == 1],
mergedframe['extweight'][mergedframe['trigA'] == 1])
effplot.update({
'pt':
numplot['pt'].divideby(denplot['pt'],
split=True,
errmethod='effnorm')
})
if stage == "B":
denplot['pt'].fill(mergedframe['pt'], mergedframe['extweight'])
numplot['pt'].fill(mergedframe['pt'][mergedframe['trigB'] == 1],
mergedframe['extweight'][mergedframe['trigB'] == 1])
effplot.update({
'pt':
numplot['pt'].divideby(denplot['pt'],
split=True,
errmethod='effnorm')
})
## Trim down events to only passing values after the pt plateu, for further studies
mergedframe = mergedframe[mergedframe['pt'] > 400]
denplot['msoft'].fill(mergedframe['msoft'], mergedframe['extweight'])
numplot['msoft'].fill(
mergedframe['msoft'][mergedframe['trigB'] == 1],
mergedframe['extweight'][mergedframe['trigB'] == 1])
effplot.update({
'msoft':
numplot['msoft'].divideby(denplot['msoft'],
split=True,
errmethod='effnorm')
})
denplot['DDBvL'].fill(mergedframe['DDBvL'], mergedframe['extweight'])
numplot['DDBvL'].fill(
mergedframe['DDBvL'][mergedframe['trigB'] == 1],
mergedframe['extweight'][mergedframe['trigB'] == 1])
effplot.update({
'DDBvL':
numplot['DDBvL'].divideby(denplot['DDBvL'],
split=True,
errmethod='effnorm')
})
if stage == "C":
tempframe = mergedframe[np.logical_and(mergedframe['s1pt'] > 140,
mergedframe['s2pt'] > 140)]
tempframe = tempframe[np.logical_and(tempframe['s1btagDeepB'] > .4184,
tempframe['s2btagDeepB'] > .4184)]
denplot['pt'].fill(tempframe['pt'], tempframe['extweight'])
numplot['pt'].fill(tempframe['pt'][mergedframe['trigC'] == 1],
tempframe['extweight'][mergedframe['trigC'] == 1])
effplot.update({
'pt':
numplot['pt'].divideby(denplot['pt'],
split=True,
errmethod='effnorm')
})
tempframe = mergedframe[mergedframe['s1pt'] > mergedframe['s2pt']]
tempframe = tempframe[np.logical_and(tempframe['s1btagDeepB'] > .4184,
tempframe['s2btagDeepB'] > .4184)]
denplot['s2pt'].fill(tempframe['s2pt'], tempframe['extweight'])
numplot['s2pt'].fill(tempframe['s2pt'][mergedframe['trigC'] == 1],
tempframe['extweight'][mergedframe['trigC'] == 1])
tempframe = mergedframe[mergedframe['s1pt'] <= mergedframe['s2pt']]
tempframe = tempframe[np.logical_and(tempframe['s1btagDeepB'] > .4184,
tempframe['s2btagDeepB'] > .4184)]
denplot['s2pt'].fill(tempframe['s1pt'], tempframe['extweight'])
numplot['s2pt'].fill(tempframe['s1pt'][mergedframe['trigC'] == 1],
tempframe['extweight'][mergedframe['trigC'] == 1])
effplot.update({
's2pt':
numplot['s2pt'].divideby(denplot['s2pt'],
split=True,
errmethod='effnorm')
})
tempframe = mergedframe[
mergedframe['s1btagDeepB'] > mergedframe['s2btagDeepB']]
tempframe = tempframe[np.logical_and(tempframe['s1pt'] > 150,
tempframe['s2pt'] > 150)]
denplot['lowb'].fill(tempframe['s2btagDeepB'], tempframe['extweight'])
numplot['lowb'].fill(
tempframe['s2btagDeepB'][mergedframe['trigC'] == 1],
tempframe['extweight'][mergedframe['trigC'] == 1])
tempframe = mergedframe[
mergedframe['s1btagDeepB'] <= mergedframe['s2btagDeepB']]
tempframe = tempframe[np.logical_and(tempframe['s1pt'] > 150,
tempframe['s2pt'] > 150)]
denplot['lowb'].fill(tempframe['s1btagDeepB'], tempframe['extweight'])
numplot['lowb'].fill(
tempframe['s1btagDeepB'][mergedframe['trigC'] == 1],
tempframe['extweight'][mergedframe['trigC'] == 1])
effplot.update({
'lowb':
numplot['lowb'].divideby(denplot['lowb'],
split=True,
errmethod='effnorm')
})
if stage == "D":
denplot['pt'].fill(mergedframe['pt'], mergedframe['extweight'])
numplot['pt'].fill(mergedframe['pt'][mergedframe['trigC'] == 1],
mergedframe['extweight'][mergedframe['trigC'] == 1])
effplot.update({
'pt':
numplot['pt'].divideby(denplot['pt'],
split=True,
errmethod='effnorm')
})
if len(stage) > 1:
if stage == "CX": stage = "C"
denplot['pt'].fill(mergedframe['pt'], mergedframe['extweight'])
numplot['pt'].fill(
mergedframe['pt'][mergedframe[f"trig{stage}"] == 1],
mergedframe['extweight'][mergedframe[f"trig{stage}"] == 1])
denplot['msoft'].fill(mergedframe['msoft'], mergedframe['extweight'])
numplot['msoft'].fill(
mergedframe['msoft'][mergedframe[f"trig{stage}"] == 1],
mergedframe['extweight'][mergedframe[f"trig{stage}"] == 1])
denplot['DDBvL'].fill(mergedframe['DDBvL'], mergedframe['extweight'])
numplot['DDBvL'].fill(
mergedframe['DDBvL'][mergedframe[f"trig{stage}"] == 1],
mergedframe['extweight'][mergedframe[f"trig{stage}"] == 1])
if "C" in stage:
tempframe = mergedframe[mergedframe['s1pt'] > mergedframe['s2pt']]
denplot['s2pt'].fill(tempframe['s2pt'], tempframe['extweight'])
numplot['s2pt'].fill(
tempframe['s2pt'][tempframe[f"trig{stage}"] == 1],
tempframe['extweight'][mergedframe[f"trig{stage}"] == 1])
tempframe = mergedframe[mergedframe['s1pt'] <= mergedframe['s2pt']]
denplot['s2pt'].fill(tempframe['s1pt'], tempframe['extweight'])
numplot['s2pt'].fill(
tempframe['s1pt'][tempframe[f"trig{stage}"] == 1],
tempframe['extweight'][mergedframe[f"trig{stage}"] == 1])
tempframe = mergedframe[
mergedframe['s1btagDeepB'] > mergedframe['s2btagDeepB']]
denplot['s2deepb'].fill(tempframe['s2btagDeepB'],
tempframe['extweight'])
numplot['s2deepb'].fill(
tempframe['s2btagDeepB'][tempframe[f"trig{stage}"] == 1],
tempframe['extweight'][mergedframe[f"trig{stage}"] == 1])
tempframe = mergedframe[
mergedframe['s1btagDeepB'] <= mergedframe['s2btagDeepB']]
denplot['s2deepb'].fill(tempframe['s1btagDeepB'],
tempframe['extweight'])
numplot['s2deepb'].fill(
tempframe['s1btagDeepB'][tempframe[f"trig{stage}"] == 1],
tempframe['extweight'][mergedframe[f"trig{stage}"] == 1])
if stage == "C": stage = "CX"
for p in denplot:
effplot.update({
p:
numplot[p].divideby(denplot[p],
split=True,
errmethod='effnorm')
})
pickle.dump(
effplot,
open(f"Seffplots/{name}_{stage}_{option}_ScaleFactor.p", 'wb'))
else:
pickle.dump(
effplot,
open(f"Effplots/{name}_EfficiencyPlot_{option}_{stage}.p", 'wb'))
for p in effplot:
effplot[p].plot(htype='err')
pickle.dump(
effplot,
open(f"Effplots/{name}_EfficiencyPlot_{option}_{stage}.p", 'wb'))
def main():
#%%################
# Plots and Setup #
###################
## Define what pdgId we expect the A to have
Aid = 9000006
## How many resolved jets we want to target with our analysis
#resjets = 4
Aid = 36
netvars = ['pt', 'eta', 'phi', 'mass', 'CSVV2', 'DeepB', 'msoft', 'DDBvL']
sigfile = 'GGH1M_Nano.root'
datafile = 'TestData.root'
bgfiles = [
'/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT200to300_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/SkimsFatJet/QCD_HT200to300_BGen_nFat1_doubB_0p5_massH_60_200_dR_2p4_18p74M.root',
'/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT300to500_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/SkimsFatJet/QCD_HT300to500_BGen_nFat1_doubB_0p5_massH_60_200_dR_2p4_17p13M.root',
'/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT500to700_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/SkimsFatJet/QCD_HT500to700_BGen_nFat1_doubB_0p5_massH_60_200_dR_2p4_8p292M.root',
'/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT700to1000_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/SkimsFatJet/QCD_HT700to1000_BGen_nFat1_doubB_0p5_massH_60_200_dR_2p4_5p845M.root',
'/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT1000to1500_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/SkimsFatJet/QCD_HT1000to1500_BGen_nFat1_doubB_0p5_massH_60_200_dR_2p4_1p953M.root',
'/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT1500to2000_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/SkimsFatJet/QCD_HT1500to2000_BGen_nFat1_doubB_0p5_massH_60_200_dR_2p4_511p5k.root',
'/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT2000toInf_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/SkimsFatJet/QCD_HT2000toInf_BGen_nFat1_doubB_0p5_massH_60_200_dR_2p4_287p3k.root'
]
#bgfiles = ['/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT200to300_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/003D724A-9341-2A40-A766-A663D3E4F10B.root',
# '/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT300to500_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/075AC8F4-7F0C-C447-82D1-CD6A47B26BCD.root',
# '/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT500to700_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/2CFFD279-3BCC-CF41-9577-B1A740DC2679.root',
# '/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT700to1000_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/0F535BA4-C750-8E44-BD53-6B3011CA2AF8.root',
# '/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT1000to1500_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/039F2902-2B95-3B4A-9EC3-53EF6299F867.root',
# '/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT1500to2000_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/71369D3B-257F-524D-A55A-55968109677A.root',
# '/cms/data/store/user/abrinke1/NanoAOD/2018/MC/QCD/QCD_HT2000toInf_BGenFilter_TuneCP5_13TeV-madgraph-pythia8/Nano25Oct2019/3542C35C-1109-D345-8B36-3DA027200467.root']
## Make a dictionary of histogram objects
# bjplots = {}
# for i in range(1,5):
# bjplots.update({
# "s_beta"+str(i): Hist(33 ,(-3.3,3.3) ,'GEN b '+str(i)+' Eta (ranked by pT)','Events','upplots/s_beta'+str(i)),
# "s_bpT"+str(i): Hist(60 ,(0,120) ,'GEN pT of b '+str(i)+' (ranked by pT)','Events','upplots/s_bpT'+str(i)),
# "s_bjetpT"+str(i): Hist(60 ,(0,120) ,'Matched RECO jet '+str(i)+' pT (ranked by b pT)','Events','upplots/s_RjetpT'+str(i)),
# "s_bjeteta"+str(i): Hist(33 ,(-3.3,3.3) ,'Matched RECO jet '+str(i)+' Eta (ranked by b pT)','Events','upplots/s_Rjeteta'+str(i)),
# "s_bjdR"+str(i): Hist(90 ,(0,3) ,'GEN b '+str(i)+' (ranked by pT) to matched jet dR','Events','upplots/s_bjdR'+str(i))
# })
plots = {
"pt":
Hist(
80, (150, 550),
'pT for highest pT jet in passing signal (red), BG (blue), and data (black) events',
'% Distribution', 'netplots/pt'),
"BGpt":
Hist(80, (150, 550)),
"SGpt":
Hist(80, (150, 550)),
"DTpt":
Hist(80, (150, 550)),
"eta":
Hist(
15, (0, 3),
'|eta| for highest pT jet in passing signal (red), BG (blue), and data (black) events',
'% Distribution', 'netplots/eta'),
"BGeta":
Hist(15, (0, 3)),
"SGeta":
Hist(15, (0, 3)),
"DTeta":
Hist(15, (0, 3)),
"phi":
Hist(
32, (-3.2, 3.2),
'phi for highest pT jet in passing signal (red), BG (blue), and data (black) events',
'% Distribution', 'netplots/phi'),
"BGphi":
Hist(32, (-3.2, 3.2)),
"SGphi":
Hist(32, (-3.2, 3.2)),
"DTphi":
Hist(32, (-3.2, 3.2)),
"mass":
Hist(
50, (0, 200),
'mass for highest pT jet in passing signal (red), BG (blue), and data (black) events',
'% Distribution', 'netplots/mass'),
"BGmass":
Hist(50, (0, 200)),
"SGmass":
Hist(50, (0, 200)),
"DTmass":
Hist(50, (0, 200)),
"CSVV2":
Hist(
22, (0, 1.1),
'CSVV2 for highest pT jet in passing signal (red), BG (blue), and data (black) events',
'% Distribution', 'netplots/CSVV2'),
"BGCSVV2":
Hist(22, (0, 1.1)),
"SGCSVV2":
Hist(22, (0, 1.1)),
"DTCSVV2":
Hist(22, (0, 1.1)),
"DeepB":
Hist(
22, (0, 1.1),
'DeepB for highest pT jet in passing signal (red), BG (blue), and data (black) events',
'% Distribution', 'netplots/DeepB'),
"BGDeepB":
Hist(22, (0, 1.1)),
"SGDeepB":
Hist(22, (0, 1.1)),
"DTDeepB":
Hist(22, (0, 1.1)),
"msoft":
Hist(
50, (0, 200),
'msoft for highest pT jet in passing signal (red), BG (blue), and data (black) events',
'% Distribution', 'netplots/msoft'),
"BGmsoft":
Hist(50, (0, 200)),
"SGmsoft":
Hist(50, (0, 200)),
"DTmsoft":
Hist(50, (0, 200)),
"DDBvL":
Hist(
22, (0, 1.1),
'DDBvL for highest pT jet in passing signal (red), BG (blue), and data (black) events',
'% Distribution', 'netplots/DDBvL'),
"BGDDBvL":
Hist(22, (0, 1.1)),
"SGDDBvL":
Hist(22, (0, 1.1)),
"DTDDBvL":
Hist(22, (0, 1.1)),
"LHEHT":
Hist(
400, (0, 4000),
'LHE_HT for highest pT jet in passing signal (red), BG (blue), and data (black) events',
'% Distribution', 'netplots/LHE_HT'),
"BGLHEHT":
Hist(400, (0, 4000)),
"SGLHEHT":
Hist(400, (0, 4000)),
"DTLHEHT":
Hist(400, (0, 4000)),
}
# for plot in bjplots:
# bjplots[plot].title = files[0]
# for plot in plots:
# plots[plot].title = files[0]
## Create an internal figure for pyplot to write to
plt.figure(1)
## Loop over input files
if True:
#####################
# Loading Variables #
#####################
print('Opening ', sigfile, ', ', datafile)
## Loop some data if the bg/signal files need to be equalized
## Open our file and grab the events tree
sigf = uproot.open(sigfile) #'nobias.root')
dataf = uproot.open(datafile)
sigevents = sigf.get('Events')
dataevents = dataf.get('Events')
bgevents = []
for bfile in bgfiles:
print('Opening ', bfile)
bgevents.append(uproot.open(bfile).get('Events'))
pdgida = sigevents.array('GenPart_pdgId')
paridxa = sigevents.array('GenPart_genPartIdxMother')
parida = pdgida[paridxa]
bs = PhysObj('bs')
## Removes all particles that do not have A parents
## from the GenPart arrays, then removes all particles
## that are not bs after resizing the pdgid array to be a valid mask
bs.oeta = pd.DataFrame(
sigevents.array('GenPart_eta')[abs(parida) == Aid][abs(pdgida)[abs(
parida) == Aid] == 5]).rename(columns=inc)
bs.ophi = pd.DataFrame(
sigevents.array('GenPart_phi')[abs(parida) == Aid][abs(pdgida)[abs(
parida) == Aid] == 5]).rename(columns=inc)
bs.opt = pd.DataFrame(
sigevents.array('GenPart_pt')[abs(parida) == Aid][abs(pdgida)[abs(
parida) == Aid] == 5]).rename(columns=inc)
## Test b order corresponds to As
testbs = pd.DataFrame(
sigevents.array('GenPart_genPartIdxMother')[abs(parida) == Aid][
abs(pdgida)[abs(parida) == Aid] == 5]).rename(columns=inc)
## The first term checks b4 has greater idx than b1, the last two check that the bs are paired
if ((testbs[4] - testbs[1]).min() <= 0) or (
(abs(testbs[2] - testbs[1]) + abs(testbs[4]) - testbs[3]).min() !=
0):
print('b to A ordering violated - time to do it the hard way')
sys.exit()
As = PhysObj('As')
As.oeta = pd.DataFrame(
sigevents.array('GenPart_eta')[abs(parida) == 25][abs(pdgida)[abs(
parida) == 25] == Aid]).rename(columns=inc)
As.ophi = pd.DataFrame(
sigevents.array('GenPart_phi')[abs(parida) == 25][abs(pdgida)[abs(
parida) == 25] == Aid]).rename(columns=inc)
As.opt = pd.DataFrame(
sigevents.array('GenPart_pt')[abs(parida) == 25][abs(pdgida)[abs(
parida) == 25] == Aid]).rename(columns=inc)
As.omass = pd.DataFrame(
sigevents.array('GenPart_mass')[abs(parida) == 25][abs(pdgida)[abs(
parida) == 25] == Aid]).rename(columns=inc)
higgs = PhysObj('higgs')
higgs.eta = pd.DataFrame(
sigevents.array('GenPart_eta')[abs(parida) != 25][abs(pdgida)[
abs(parida) != 25] == 25]).rename(columns=inc)
higgs.phi = pd.DataFrame(
sigevents.array('GenPart_phi')[abs(parida) != 25][abs(pdgida)[
abs(parida) != 25] == 25]).rename(columns=inc)
higgs.pt = pd.DataFrame(
sigevents.array('GenPart_pt')[abs(parida) != 25][abs(pdgida)[
abs(parida) != 25] == 25]).rename(columns=inc)
sigjets = PhysObj('sigjets')
sigjets.eta = np.abs(
pd.DataFrame(sigevents.array('FatJet_eta')).rename(columns=inc))
sigjets.phi = pd.DataFrame(
sigevents.array('FatJet_phi')).rename(columns=inc)
sigjets.pt = pd.DataFrame(
sigevents.array('FatJet_pt')).rename(columns=inc)
sigjets.mass = pd.DataFrame(
sigevents.array('FatJet_mass')).rename(columns=inc)
sigjets.CSVV2 = pd.DataFrame(
sigevents.array('FatJet_btagCSVV2')).rename(columns=inc)
sigjets.DeepB = pd.DataFrame(
sigevents.array('FatJet_btagDeepB')).rename(columns=inc)
sigjets.DDBvL = pd.DataFrame(
sigevents.array('FatJet_btagDDBvL')).rename(columns=inc)
sigjets.msoft = pd.DataFrame(
sigevents.array('FatJet_msoftdrop')).rename(columns=inc)
sigjets.LHEHT = pd.DataFrame(
sigevents.array('LHE_HT')).rename(columns=inc)
slimjets = PhysObj('slimjets')
slimjets.eta = pd.DataFrame(
sigevents.array('Jet_eta')).rename(columns=inc)
slimjets.phi = pd.DataFrame(
sigevents.array('Jet_phi')).rename(columns=inc)
slimjets.pt = pd.DataFrame(
sigevents.array('Jet_pt')).rename(columns=inc)
slimjets.mass = pd.DataFrame(
sigevents.array('Jet_mass')).rename(columns=inc)
#sigjets.CSVV2 = pd.DataFrame(sigevents.array('FatJet_btagCSVV2')).rename(columns=inc)
slimjets.DeepB = pd.DataFrame(
sigevents.array('Jet_btagDeepB')).rename(columns=inc)
#sigjets.DDBvL = pd.DataFrame(sigevents.array('FatJet_btagDDBvL')).rename(columns=inc)
#sigjets.msoft = pd.DataFrame(sigevents.array('FatJet_msoftdrop')).rename(columns=inc)
slimjets.DeepFB = pd.DataFrame(
sigevents.array('Jet_btagDeepFlavB')).rename(columns=inc)
slimjets.puid = pd.DataFrame(
sigevents.array('Jet_puId')).rename(columns=inc)
datajets = PhysObj('datajets')
datajets.eta = np.abs(
pd.DataFrame(dataevents.array('FatJet_eta')).rename(columns=inc))
datajets.phi = pd.DataFrame(
dataevents.array('FatJet_phi')).rename(columns=inc)
datajets.pt = pd.DataFrame(
dataevents.array('FatJet_pt')).rename(columns=inc)
datajets.mass = pd.DataFrame(
dataevents.array('FatJet_mass')).rename(columns=inc)
datajets.CSVV2 = pd.DataFrame(
dataevents.array('FatJet_btagCSVV2')).rename(columns=inc)
datajets.DeepB = pd.DataFrame(
dataevents.array('FatJet_btagDeepB')).rename(columns=inc)
datajets.DDBvL = pd.DataFrame(
dataevents.array('FatJet_btagDDBvL')).rename(columns=inc)
datajets.msoft = pd.DataFrame(
dataevents.array('FatJet_msoftdrop')).rename(columns=inc)
datajets.LHEHT = pd.DataFrame(
sigevents.array('LHE_HT')).rename(columns=inc)
#bgjets.DeepFB= pd.DataFrame(bgevents.array('Jet_btagDeepFlavB')).rename(columns=inc)
bgjets = [
PhysObj('300'),
PhysObj('500'),
PhysObj('700'),
PhysObj('1000'),
PhysObj('1500'),
PhysObj('2000'),
PhysObj('inf')
]
for i in range(7):
bgjets[i].eta = np.abs(
pd.DataFrame(
bgevents[i].array('FatJet_eta')).rename(columns=inc))
bgjets[i].phi = pd.DataFrame(
bgevents[i].array('FatJet_phi')).rename(columns=inc)
bgjets[i].pt = pd.DataFrame(
bgevents[i].array('FatJet_pt')).rename(columns=inc)
bgjets[i].mass = pd.DataFrame(
bgevents[i].array('FatJet_mass')).rename(columns=inc)
bgjets[i].CSVV2 = pd.DataFrame(
bgevents[i].array('FatJet_btagCSVV2')).rename(columns=inc)
bgjets[i].DeepB = pd.DataFrame(
bgevents[i].array('FatJet_btagDeepB')).rename(columns=inc)
bgjets[i].DDBvL = pd.DataFrame(
bgevents[i].array('FatJet_btagDDBvL')).rename(columns=inc)
bgjets[i].msoft = pd.DataFrame(
bgevents[i].array('FatJet_msoftdrop')).rename(columns=inc)
bgjets[i].LHEHT = pd.DataFrame(
bgevents[i].array('LHE_HT')).rename(columns=inc)
del bgevents
print('Processing ' + str(len(bs.oeta)) + ' events')
## Figure out how many bs and jets there are
nb = bs.oeta.shape[1]
njet = sigjets.eta.shape[1]
#nsjet=slimjets.eta.shape[1]
na = As.oeta.shape[1]
if na != 2:
print("More than two As per event, found " + str(na) + ", halting")
sys.exit()
## Create sorted versions of A values by pt
for prop in ['eta', 'phi', 'pt', 'mass']:
As[prop] = pd.DataFrame()
for i in range(1, 3):
As[prop][i] = As['o' + prop][As.opt.rank(
axis=1, ascending=False, method='first') == i].max(axis=1)
## Clean up original ordered dataframes; we don't really need them
#del As['o'+prop]
## Reorder out b dataframes to match sorted A parents
tframe = pd.DataFrame()
tframe[1] = (As.opt.rank(axis=1, ascending=False,
method='first') == 1)[1]
tframe[2] = (As.opt.rank(axis=1, ascending=False,
method='first') == 1)[1]
tframe[3] = (As.opt.rank(axis=1, ascending=False,
method='first') == 1)[2]
tframe[4] = (As.opt.rank(axis=1, ascending=False,
method='first') == 1)[2]
for prop in ['eta', 'phi', 'pt']:
bs[prop] = pd.DataFrame()
bs[prop][1] = bs['o' + prop][tframe][1].dropna().append(
bs['o' + prop][tframe][3].dropna()).sort_index()
bs[prop][2] = bs['o' + prop][tframe][2].dropna().append(
bs['o' + prop][tframe][4].dropna()).sort_index()
bs[prop][3] = bs['o' + prop][~tframe][1].dropna().append(
bs['o' + prop][~tframe][3].dropna()).sort_index()
bs[prop][4] = bs['o' + prop][~tframe][2].dropna().append(
bs['o' + prop][~tframe][4].dropna()).sort_index()
## Clean up original ordered dataframes; we don't really need them.
#del bs['o'+prop]
# ## Sort our b dataframes in descending order of pt
# for prop in ['spt','seta','sphi']:
# bs[prop] = pd.DataFrame()
# #bs.spt, bs.seta, bs.sphi = pd.DataFrame(), pd.DataFrame(), pd.DataFrame()
# for i in range(1,nb+1):
# bs[prop][i] = bs[prop[1:]][bs.pt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
# #bs.seta[i] = bs.eta[bs.pt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
# #bs.sphi[i] = bs.phi[bs.pt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
# plots['genAmass'].dfill(As.mass)
ev = Event(bs, sigjets, As, higgs)
for jets in [sigjets, datajets]:
jets.cut(jets.pt > 170)
jets.cut(abs(jets.eta) < 2.4)
jets.cut(jets.DDBvL > 0.6)
jets.cut(jets.DeepB > 0.4184)
jets.cut(jets.msoft > 0.25)
for jets in bgjets:
#pass
jets.cut(jets.pt > 170)
jets.cut(abs(jets.eta) < 2.4)
jets.cut(jets.DDBvL > 0.6)
jets.cut(jets.DeepB > 0.4184)
jets.cut(jets.msoft > 0.25)
bs.cut(bs.pt > 5)
bs.cut(abs(bs.eta) < 2.4)
ev.sync()
slimjets.cut(slimjets.DeepB > 0.1241)
slimjets.cut(slimjets.DeepFB > 0.277)
slimjets.cut(slimjets.puid > 0)
slimjets.trimTo(jets.eta)
##############################
# Processing and Calculation #
##############################
## Create our dR dataframe by populating its first column and naming it accordingly
jbdr2 = pd.DataFrame(
np.power(sigjets.eta[1] - bs.eta[1], 2) +
np.power(sigjets.phi[1] - bs.phi[1], 2)).rename(
columns={1: 'Jet 1 b 1'})
sjbdr2 = pd.DataFrame(
np.power(slimjets.eta[1] - bs.eta[1], 2) +
np.power(slimjets.phi[1] - bs.phi[1], 2)).rename(
columns={1: 'Jet 1 b 1'})
## Loop over jet x b combinations
jbstr = []
for j in range(1, njet + 1):
for b in range(1, nb + 1):
## Make our column name
jbstr.append("Jet " + str(j) + " b " + str(b))
if (j + b == 2):
continue
## Compute and store the dr of the given b and jet for every event at once
jbdr2[jbstr[-1]] = pd.DataFrame(
np.power(sigjets.eta[j] - bs.eta[b], 2) +
np.power(sigjets.phi[j] - bs.phi[b], 2))
sjbdr2[jbstr[-1]] = pd.DataFrame(
np.power(slimjets.eta[j] - bs.eta[b], 2) +
np.power(slimjets.phi[j] - bs.phi[b], 2))
## Create a copy array to collapse in jets instead of bs
blist = []
sblist = []
for b in range(nb):
blist.append(np.sqrt(jbdr2.filter(like='b ' + str(b + 1))))
blist[b] = blist[b][blist[b].rank(axis=1, method='first') == 1]
blist[b] = blist[b].rename(columns=lambda x: int(x[4:6]))
sblist.append(np.sqrt(sjbdr2.filter(like='b ' + str(b + 1))))
sblist[b] = sblist[b][sblist[b].rank(axis=1, method='first') == 1]
sblist[b] = sblist[b].rename(columns=lambda x: int(x[4:6]))
## Trim resolved jet objects
# if resjets==3:
# for i in range(nb):
# for j in range(nb):
# if i != j:
# blist[i] = blist[i][np.logical_not(blist[i] > blist[j])]
# blist[i] = blist[i][blist[i]<0.4]
## Cut our events to only events with 3-4 bs in one fatjet of dR<0.8
fjets = blist[0][blist[0] < 0.8].fillna(0) / blist[0][
blist[0] < 0.8].fillna(0)
for i in range(1, 4):
fjets = fjets + blist[i][blist[i] < 0.8].fillna(0) / blist[i][
blist[i] < 0.8].fillna(0)
fjets = fjets.max(axis=1)
fjets = fjets[fjets == 4].dropna()
sigjets.trimTo(fjets)
ev.sync()
#
#######################
# Training Neural Net #
#######################
datajetframe = pd.DataFrame()
#for i in range(njet):
for prop in netvars:
datajetframe[prop] = datajets[prop][datajets['pt'].rank(
axis=1, method='first') == 1].max(axis=1)
#bgjetframe['val'] = 0
datajetframe['LHEHT'] = datajets['LHEHT']
sigjetframe = pd.DataFrame()
#for i in range(njet):
for prop in netvars:
sigjetframe[prop] = sigjets[prop][sigjets['pt'].rank(
axis=1, method='first') == 1].max(axis=1)
#sigjetframe['val'] = 1
sigjetframe['LHEHT'] = sigjets['LHEHT']
#X_train = pd.concat([bgjetframe.sample(frac=0.7,random_state=6),sigjetframe.sample(frac=0.7,random_state=6)])
print('Signal cut to', sigjetframe.shape[0], ' events')
print('Data has', datajetframe.shape[0], 'events')
bgweights = [1, 0.259, 0.0515, 0.01666, 0.00905, 0.003594, 0.001401]
bgpieces = []
for i in range(len(bgweights)):
tempframe = pd.DataFrame()
#for i in range(njet):
for prop in netvars:
tempframe[prop] = bgjets[i][prop][bgjets[i]['pt'].rank(
axis=1, method='first') == 1].max(axis=1)
tempframe['LHEHT'] = bgjets[i]['LHEHT']
tempframe = tempframe.sample(frac=bgweights[i], random_state=6)
bgpieces.append(tempframe)
bgjetframe = pd.concat(bgpieces)
bgjetframe = bgjetframe[bgjetframe != 0]
bgjetframe = bgjetframe.dropna()
print('Background has', bgjetframe.shape[0], 'events')
del bgpieces
netvars.append('LHEHT')
for col in netvars:
plots['BG' + col].fill(bgjetframe[col])
plots['SG' + col].fill(sigjetframe[col])
plots['DT' + col].fill(datajetframe[col])
plt.clf()
for col in netvars:
plots['BG' +
col][0] = plots['BG' + col][0] / sum(plots['BG' + col][0])
plots['SG' +
col][0] = plots['SG' + col][0] / sum(plots['SG' + col][0])
#print(col)
#print(plots['DT'+col][0])
#print(datajetframe[col])
plots['DT' +
col][0] = plots['DT' + col][0] / sum(plots['DT' + col][0])
##p.norm(sum(p[0]))
#p[0] = p[0]/sum(p[0])
for col in netvars:
plt.clf()
plots['SG' + col].make(color='red', linestyle='-', htype='step')
plots['DT' + col].make(color='black', linestyle='--', htype='step')
plots['BG' + col].make(color='blue', linestyle=':', htype='step')
plots[col].plot(same=True)
def ana(files,returnplots=False):
#%%################
# Plots and Setup #
###################
## Define what pdgId we expect the A to have
Aid = 9000006
## How many resolved jets we want to target with our analysis
resjets = 4
#Aid = 36
## Make a dictionary of histogram objects
bjplots = {}
for i in range(1,5):
bjplots.update({
"s_beta"+str(i): Hist(33 ,(-3.3,3.3) ,'GEN b '+str(i)+' Eta (ranked by pT)','Events','upplots/s_beta'+str(i)),
"s_bpT"+str(i): Hist(60 ,(0,120) ,'GEN pT of b '+str(i)+' (ranked by pT)','Events','upplots/s_bpT'+str(i)),
"s_bjetpT"+str(i): Hist(60 ,(0,120) ,'Matched RECO jet '+str(i)+' pT (ranked by b pT)','Events','upplots/s_RjetpT'+str(i)),
"s_bjeteta"+str(i): Hist(33 ,(-3.3,3.3) ,'Matched RECO jet '+str(i)+' Eta (ranked by b pT)','Events','upplots/s_Rjeteta'+str(i)),
"s_bjdR"+str(i): Hist(90 ,(0,3) ,'GEN b '+str(i)+' (ranked by pT) to matched jet dR','Events','upplots/s_bjdR'+str(i))
})
plots = {
"HpT": Hist(60 ,(0,320) ,'GEN Higgs pT','Events','upplots/HpT'),
#"HAdR": Hist(100,(0,2) ,'GEN Higgs to A dR','Events','upplots/HAdR'),
#'HAdeta': Hist(66 ,(-3.3,3.3) ,'GEN Higgs to A deta','Events','upplots/HAdeta'),
#'HAdphi': Hist(66 ,(-3.3,3.3) ,'GEN Higgs to A dphi','Events','upplots/HAdphi'),
"A1pT": Hist(80 ,(0,160) ,'Highest GEN A pT','Events','upplots/A1pT'),
"A2pT": Hist(80 ,(0,160) ,'Lowest GEN A pT','Events','upplots/A2pT'),
"AdR": Hist(50 ,(0,5) ,'GEN A1 to A2 dR','Events','upplots/AdR'),
"bdRA1": Hist(50 ,(0,5) ,'GEN dR between highest pT A child bs','Events','upplots/bdRA1'),
"bdRA2": Hist(50 ,(0,5) ,'GEN dR between lowest pT A child bs','Events','upplots/bdRA2'),
"bdetaA1": Hist(34 ,(0,3.4) ,'GEN |deta| between highest-A child bs','Events','upplots/bdetaA1'),
"bdetaA2": Hist(34 ,(0,3.4) ,'GEN |deta| between lowest-A child bs','Events','upplots/bdetaA2'),
"bdphiA1": Hist(34 ,(0,3.4) ,'GEN |dphi| between highest-A child bs','Events','upplots/bdphiA1'),
"bdphiA2": Hist(34 ,(0,3.4) ,'GEN |dphi| between lowest-A child bs','Events','upplots/bdphiA2'),
"bphi": Hist(66 ,(-3.3,3.3) ,'GEN b Phi','Events','upplots/bphi'),
"bjdR": Hist(100,(0,2) ,'All GEN bs to matched jet dR','Events','upplots/bjdR'),
"RjetpT": Hist(100,(0,100) ,'RECO matched jet pT','Events','upplots/RjetpT'),
"Rjeteta": Hist(66 ,(-3.3,3.3) ,'RECO matched jet eta','Events','upplots/Rjeteta'),
"RjetCSVV2":Hist(140 ,(-12,2) ,'RECO matched jet btagCSVV2 score','events','upplots/RjetCSVV2'),
"RjetDeepB":Hist(40 ,(-2.5,1.5) ,'RECO matched jet btagDeepB score','events','upplots/RjetDeepB'),
"RjetDeepFB" :Hist(24 ,(0,1.2) ,'RECO matched jet btagDeepFlavB score','events','upplots/RjetDeepFB'),
"RA1pT": Hist(80 ,(0,160) ,'pT of RECO A1 objects constructed from matched jets','Events','upplots/RA1pT'),
"RA2pT": Hist(80 ,(0,160) ,'pT of RECO A2 objects constructed from matched jets','Events','upplots/RA2pT'),
"RA1mass": Hist(40 ,(0,80) ,'reconstructed mass of A1 objects from matched jets','Events','upplots/RA1mass'),
"RA2mass": Hist(40 ,(0,80) ,'reconstructed mass of A2 objects from matched jets','Events','upplots/RA2mass'),
"RA1dR": Hist(50 ,(0,5) ,'dR between jet children of reconstructed A1 object','Events','upplots/RA1dR'),
"RA2dR": Hist(50 ,(0,5) ,'dR between jet children of reconstructed A2 object','Events','upplots/RA2dR'),
"RA1deta": Hist(33 ,(0,3.3) ,'|deta| between jet children of reconstructed A1 object','Events','upplots/RA1deta'),
"RA2deta": Hist(33 ,(0,3.3) ,'|deta| between jet children of reconstructed A2 object','Events','upplots/RA2deta'),
"RA1dphi": Hist(33 ,(0,3.3) ,'|dphi| between jet children of reconstructed A1 object','Events','upplots/RA1dphi'),
"RA2dphi": Hist(33 ,(0,3.3) ,'|dphi| between jet children of reconstructed A2 object','Events','upplots/RA2dphi'),
"RHmass": Hist(80 ,(0,160) ,'reconstructed mass of Higgs object from reconstructed As','Events','upplots/RHmass'),
"RHpT": Hist(100,(0,200) ,'pT of reconstructed higgs object from reconstructed As','Events','upplots/RHpT'),
"RHdR": Hist(50 ,(0,5) ,'dR between A children of reconstructed higgs object','Events','upplots/RHdR'),
"RHdeta": Hist(33 ,(0,3.3) ,'|deta| between A children of reconstructed higgs object','Events','upplots/RHdeta'),
"RHdphi": Hist(33 ,(0,3.3) ,'|dphi| between A children of reconstructed higgs object','Events','upplots/RHdphi'),
##
"RalljetpT": Hist(100,(0,100),'All RECO jet pT','Events','upplots/RalljetpT'),
"bjdRvlogbpT1": Hist2d([80,200],[[0,8],[0,2]],'log2(GEN b pT)','dR from 1st pT GEN b to matched RECO jet','upplots/bjdRvlogbpT1'),
"bjdRvlogbpT2": Hist2d([80,200],[[0,8],[0,2]],'log2(GEN b pT)','dR from 2nd pT GEN b to matched RECO jet','upplots/bjdRvlogbpT2'),
"bjdRvlogbpT3": Hist2d([80,200],[[0,8],[0,2]],'log2(GEN b pT)','dR from 3rd pT GEN b to matched RECO jet','upplots/bjdRvlogbpT3'),
"bjdRvlogbpT4": Hist2d([80,200],[[0,8],[0,2]],'log2(GEN b pT)','dR from 4th pT GEN b to matched RECO jet','upplots/bjdRvlogbpT4'),
"jetoverbpTvlogbpT1": Hist2d([60,40],[[2,8],[0,4]],'log2(GEN b pT)','RECO jet pT / 1st GEN b pT for matched jets','upplots/jetoverbpTvlogbpT1'),
"jetoverbpTvlogbpT2": Hist2d([60,40],[[2,8],[0,4]],'log2(GEN b pT)','RECO jet pT / 2nd GEN b pT for matched jets','upplots/jetoverbpTvlogbpT2'),
"jetoverbpTvlogbpT3": Hist2d([60,40],[[2,8],[0,4]],'log2(GEN b pT)','RECO jet pT / 3rd GEN b pT for matched jets','upplots/jetoverbpTvlogbpT3'),
"jetoverbpTvlogbpT4": Hist2d([60,40],[[2,8],[0,4]],'log2(GEN b pT)','RECO jet pT / 4th GEN b pT for matched jets','upplots/jetoverbpTvlogbpT4'),
"npassed": Hist(1 ,(0.5,1.5) ,'','Number of events that passed cuts', 'upplots/npassed'),
"genAmass": Hist(40 ,(0,80) ,'GEN mass of A objects','Events','upplots/Amass_g'),
"cutAmass": Hist(40 ,(0,80) ,'GEN mass of A objects that pass cuts','Events','upplots/Amass_c')
}
for plot in bjplots:
bjplots[plot].title = files[0]
for plot in plots:
plots[plot].title = files[0]
## Create an internal figure for pyplot to write to
plt.figure(1)
## Loop over input files
for fnum in range(len(files)):
#####################
# Loading Variables #
#####################
print('Opening '+files[fnum])
## Open our file and grab the events tree
f = uproot.open(files[fnum])#'nobias.root')
events = f.get('Events')
pdgida = events.array('GenPart_pdgId')
paridxa = events.array('GenPart_genPartIdxMother')
parida = pdgida[paridxa]
bs = PhysObj('bs')
## Removes all particles that do not have A parents
## from the GenPart arrays, then removes all particles
## that are not bs after resizing the pdgid array to be a valid mask
bs.oeta = pd.DataFrame(events.array('GenPart_eta')[abs(parida)==Aid][abs(pdgida)[abs(parida)==Aid]==5]).rename(columns=inc)
bs.ophi = pd.DataFrame(events.array('GenPart_phi')[abs(parida)==Aid][abs(pdgida)[abs(parida)==Aid]==5]).rename(columns=inc)
bs.opt = pd.DataFrame(events.array('GenPart_pt' )[abs(parida)==Aid][abs(pdgida)[abs(parida)==Aid]==5]).rename(columns=inc)
## Test b order corresponds to As
testbs = pd.DataFrame(events.array('GenPart_genPartIdxMother')[abs(parida)==Aid][abs(pdgida)[abs(parida)==Aid]==5]).rename(columns=inc)
## The first term checks b4 has greater idx than b1, the last two check that the bs are paired
if ((testbs[4]-testbs[1]).min() <= 0) or ((abs(testbs[2]-testbs[1]) + abs(testbs[4])-testbs[3]).min() != 0):
print('b to A ordering violated - time to do it the hard way')
sys.exit()
As = PhysObj('As')
As.oeta = pd.DataFrame(events.array('GenPart_eta')[abs(parida)==25][abs(pdgida)[abs(parida)==25]==Aid]).rename(columns=inc)
As.ophi = pd.DataFrame(events.array('GenPart_phi')[abs(parida)==25][abs(pdgida)[abs(parida)==25]==Aid]).rename(columns=inc)
As.opt = pd.DataFrame(events.array('GenPart_pt' )[abs(parida)==25][abs(pdgida)[abs(parida)==25]==Aid]).rename(columns=inc)
As.omass =pd.DataFrame(events.array('GenPart_mass')[abs(parida)==25][abs(pdgida)[abs(parida)==25]==Aid]).rename(columns=inc)
higgs = PhysObj('higgs')
higgs.eta = pd.DataFrame(events.array('GenPart_eta')[abs(parida)!=25][abs(pdgida)[abs(parida)!=25]==25]).rename(columns=inc)
higgs.phi = pd.DataFrame(events.array('GenPart_phi')[abs(parida)!=25][abs(pdgida)[abs(parida)!=25]==25]).rename(columns=inc)
higgs.pt = pd.DataFrame(events.array('GenPart_pt' )[abs(parida)!=25][abs(pdgida)[abs(parida)!=25]==25]).rename(columns=inc)
jets = PhysObj('jets')
jets.eta= pd.DataFrame(events.array('Jet_eta')).rename(columns=inc)
jets.phi= pd.DataFrame(events.array('Jet_phi')).rename(columns=inc)
jets.pt = pd.DataFrame(events.array('Jet_pt')).rename(columns=inc)
jets.mass=pd.DataFrame(events.array('Jet_mass')).rename(columns=inc)
jets.CSVV2 = pd.DataFrame(events.array('Jet_btagCSVV2')).rename(columns=inc)
jets.DeepB = pd.DataFrame(events.array('Jet_btagDeepB')).rename(columns=inc)
jets.DeepFB= pd.DataFrame(events.array('Jet_btagDeepFlavB')).rename(columns=inc)
print('Processing ' + str(len(bs.oeta)) + ' events')
## Figure out how many bs and jets there are
nb = bs.oeta.shape[1]
njet= jets.eta.shape[1]
na = As.oeta.shape[1]
if na != 2:
print("More than two As per event, found "+str(na)+", halting")
sys.exit()
## Create sorted versions of A values by pt
for prop in ['eta','phi','pt','mass']:
As[prop] = pd.DataFrame()
for i in range(1,3):
As[prop][i] = As['o'+prop][As.opt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
## Clean up original ordered dataframes; we don't really need them
#del As['o'+prop]
## Reorder out b dataframes to match sorted A parents
tframe = pd.DataFrame()
tframe[1] = (As.opt.rank(axis=1,ascending=False,method='first')==1)[1]
tframe[2] = (As.opt.rank(axis=1,ascending=False,method='first')==1)[1]
tframe[3] = (As.opt.rank(axis=1,ascending=False,method='first')==1)[2]
tframe[4] = (As.opt.rank(axis=1,ascending=False,method='first')==1)[2]
for prop in ['eta','phi','pt']:
bs[prop] = pd.DataFrame()
bs[prop][1] = bs['o'+prop][tframe][1].dropna().append(bs['o'+prop][tframe][3].dropna()).sort_index()
bs[prop][2] = bs['o'+prop][tframe][2].dropna().append(bs['o'+prop][tframe][4].dropna()).sort_index()
bs[prop][3] = bs['o'+prop][~tframe][1].dropna().append(bs['o'+prop][~tframe][3].dropna()).sort_index()
bs[prop][4] = bs['o'+prop][~tframe][2].dropna().append(bs['o'+prop][~tframe][4].dropna()).sort_index()
## Clean up original ordered dataframes; we don't really need them.
#del bs['o'+prop]
## Sort our b dataframes in descending order of pt
for prop in ['spt','seta','sphi']:
bs[prop] = pd.DataFrame()
#bs.spt, bs.seta, bs.sphi = pd.DataFrame(), pd.DataFrame(), pd.DataFrame()
for i in range(1,nb+1):
bs[prop][i] = bs[prop[1:]][bs.pt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
#bs.seta[i] = bs.eta[bs.pt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
#bs.sphi[i] = bs.phi[bs.pt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
plots['genAmass'].dfill(As.mass)
ev = Event(bs,jets,As,higgs)
jets.cut(jets.pt>0)
bs.cut(bs.pt>0)
ev.sync()
##############################
# Processing and Calculation #
##############################
## Create our dR dataframe by populating its first column and naming it accordingly
jbdr2 = pd.DataFrame(np.power(jets.eta[1]-bs.eta[1],2) + np.power(jets.phi[1]-bs.phi[1],2)).rename(columns={1:'Jet 1 b 1'})
sjbdr2= pd.DataFrame(np.power(jets.eta[1]-bs.seta[1],2) + np.power(jets.phi[1]-bs.sphi[1],2)).rename(columns={1:'Jet 1 b 1'})
## Loop over jet x b combinations
jbstr = []
for j in range(1,njet+1):
for b in range(1,nb+1):
## Make our column name
jbstr.append("Jet "+str(j)+" b "+str(b))
if (j+b==2):
continue
## Compute and store the dr of the given b and jet for every event at once
jbdr2[jbstr[-1]] = pd.DataFrame(np.power(jets.eta[j]-bs.eta[b],2) + np.power(jets.phi[j]-bs.phi[b],2))
sjbdr2[jbstr[-1]]= pd.DataFrame(np.power(jets.eta[j]-bs.seta[b],2) + np.power(jets.phi[j]-bs.sphi[b],2))
## Create a copy array to collapse in jets instead of bs
blist = []
sblist = []
for b in range(nb):
blist.append(np.sqrt(jbdr2.filter(like='b '+str(b+1))))
blist[b] = blist[b][blist[b].rank(axis=1,method='first') == 1]
blist[b] = blist[b].rename(columns=lambda x:int(x[4:6]))
sblist.append(np.sqrt(sjbdr2.filter(like='b '+str(b+1))))
sblist[b] = sblist[b][sblist[b].rank(axis=1,method='first') == 1]
sblist[b] = sblist[b].rename(columns=lambda x:int(x[4:6]))
## Trim resolved jet objects
if resjets==3:
for i in range(nb):
for j in range(nb):
if i != j:
blist[i] = blist[i][np.logical_not(blist[i] > blist[j])]
blist[i] = blist[i][blist[i]<0.4]
## Cut our events to only resolved 4jet events with dR<0.4
rjets = blist[0][blist[0]<0.4].fillna(0)
for i in range(1,4):
rjets = np.logical_or(rjets,blist[i][blist[i]<0.4].fillna(0))
rjets = rjets.sum(axis=1)
rjets = rjets[rjets==resjets].dropna()
jets.trimTo(rjets)
ev.sync()
#############################
# Constructing RECO objects #
#############################
for prop in ['bpt','beta','bphi','bmass']:
jets[prop] = pd.DataFrame()
for i in range(nb):
jets[prop][i+1] = jets[prop[1:]][blist[i]>0].max(axis=1)
jets.bdr = pd.DataFrame()
for i in range(nb):
jets.bdr[i+1] = blist[i][blist[i]>0].max(axis=1)
ev.sync()
if resjets==3:
pidx = [2,1,4,3]
for prop in ['bpt','beta','bphi','bmass']:
jets[prop]['merged'], jets[prop]['missing'] = (jets[prop][1]==jets[prop][3]),(jets[prop][1]==jets[prop][3])
for i in range(1,nb+1):
jets[prop]['merged']=jets[prop]['merged']+jets[prop].fillna(0)[i][(jets.bmass[i]>=15) & (jets.bmass[i]+jets.bmass[pidx[i-1]]==jets.bmass[i])]
jets[prop]['missing']=jets[prop]['missing']+jets[prop].fillna(0)[i][(jets.bmass[i]<15) & (jets.bmass[i]+jets.bmass[pidx[i-1]]==jets.bmass[i])]
#jets[prop][i] = jets[prop][i]+(0*jets[prop][pidx])
bvec = []
for i in range(1,nb+1):
bvec.append(TLorentzVectorArray.from_ptetaphim(jets.bpt[i],jets.beta[i],jets.bphi[i],jets.bmass[i]))
avec = []
for i in range(0,nb,2):
avec.append(bvec[i]+bvec[i+1])
for prop in ['apt','aeta','aphi','amass']:
jets[prop] = pd.DataFrame()
for i in range(na):
jets.apt[i+1] = avec[i].pt
jets.aeta[i+1] = avec[i].eta
jets.aphi[i+1] = avec[i].phi
jets.amass[i+1]= avec[i].mass
for prop in ['apt','aeta','aphi','amass']:
jets[prop].index = jets.pt.index
hvec = [avec[0]+avec[1]]
for prop in ['hpt','heta','hphi','hmass']:
jets[prop] = pd.DataFrame()
jets.hpt[1] = hvec[0].pt
jets.heta[1] = hvec[0].eta
jets.hphi[1] = hvec[0].phi
jets.hmass[1]= hvec[0].mass
for prop in ['hpt','heta','hphi','hmass']:
jets[prop].index = jets.eta.index
################
# Filling Data #
################
for i in range(4):
plots['bjdRvlogbpT'+str(i+1)].dfill(np.log2(bs.spt[[i+1]]),bs.trim(sblist[i]))
plots['RjetCSVV2'].dfill(jets.CSVV2[blist[i]>0])
plots['RjetDeepB'].dfill(jets.DeepB[blist[i]>0])
plots['RjetDeepFB'].dfill(jets.DeepFB[blist[i]>0])
yval = np.divide(jets.pt[sblist[i]>0].melt(value_name=0).drop('variable',axis=1).dropna().reset_index(drop=True)[0],bs.spt[[i+1]].dropna().reset_index(drop=True)[i+1])
xval = np.log2(bs.spt[[i+1]]).melt(value_name=0).drop('variable',axis=1).dropna().reset_index(drop=True)[0]
plots['jetoverbpTvlogbpT'+str(i+1)].fill(xval,yval)
bjplots['s_bpT'+str(i+1)].dfill(bs.spt[[i+1]])
bjplots['s_beta'+str(i+1)].dfill(bs.seta[[i+1]])
bjplots['s_bjetpT'+str(i+1)].dfill(jets.pt[sblist[i]>0])
bjplots['s_bjeteta'+str(i+1)].dfill(jets.eta[sblist[i]>0])
bjplots['s_bjdR'+str(i+1)].dfill(sblist[i][sblist[i]!=0])
plots['HpT'].dfill(higgs.pt)
plots['A1pT'].fill(As.pt[1])
plots['A2pT'].fill(As.pt[2])
plots['AdR'].fill(np.sqrt(np.power(As.eta[2]-As.eta[1],2) + np.power(As.phi[2]-As.phi[1],2)))
plots['bdRA1'].fill(np.sqrt(np.power(bs.eta[2]-bs.eta[1],2) + np.power(bs.phi[2]-bs.phi[1],2)))
plots['bdRA2'].fill(np.sqrt(np.power(bs.eta[4]-bs.eta[3],2) + np.power(bs.phi[4]-bs.phi[3],2)))
plots['bdetaA1'].fill(abs(bs.eta[2]-bs.eta[1]))
plots['bdetaA2'].fill(abs(bs.eta[4]-bs.eta[3]))
plots['bdphiA1'].fill(abs(bs.phi[2]-bs.phi[1]))
plots['bdphiA2'].fill(abs(bs.phi[4]-bs.phi[3]))
plots['bphi'].dfill(bs.phi)
plots['RalljetpT'].dfill(jets.pt)
plots['bjdR'].dfill(jets.bdr)
plots['RjetpT'].dfill(jets.bpt)
plots['Rjeteta'].dfill(jets.beta)
for i in range(1,3):
plots['RA'+str(i)+'pT' ].fill(jets.apt[i])
plots['RA'+str(i)+'mass'].fill(jets.amass[i])
#lots['RA'+str(i)+'deta'].fill(abs(jets.beta[2*i]-jets.beta[(2*i)-1]))
plots['RA'+str(i)+'dR' ].fill(np.sqrt(np.power(jets.beta[2*i]-jets.beta[(2*i)-1],2)+np.power(jets.bphi[2*i]-jets.bphi[(2*i)-1],2)))
plots['RA'+str(i)+'deta'].fill(abs(jets.beta[2*i]-jets.beta[(2*i)-1]))
plots['RA'+str(i)+'dphi'].fill(abs(jets.bphi[2*i]-jets.bphi[(2*i)-1]))
plots['RHpT' ].fill(jets.hpt[1])
plots['RHmass'].fill(jets.hmass[1])
plots['RHdR' ].fill(np.sqrt(np.power(jets.aeta[2]-jets.aeta[1],2)+np.power(jets.aphi[2]-jets.aphi[1],2)))
plots['RHdeta'].fill(abs(jets.aeta[2]-jets.aeta[1]))
plots['RHdphi'].fill(abs(jets.aphi[2]-jets.aphi[1]))
plots['npassed'].fill(jets.hpt[1]/jets.hpt[1])
plots['cutAmass'].dfill(As.mass)
############
# Plotting #
############
plt.clf()
#plots.pop('bjdR').plot(logv=True)
for i in range(1,5):
bjplots.pop('s_bjdR'+str(i)).plot(logv=True)
for p in plots:
plots[p].plot()
for p in bjplots:
bjplots[p].plot()
#%%
if returnplots==True:
return plots
else:
sys.exit()
def trig(files):
## Create a dictionary of histogram objects
rptbins = [0,0.5,1,1.5,2,2.5,3,3.5,4,4.5,5,5.5,6,6.5,7,7.5,8,8.5,9,9.5,10,11,12,13,14,15,30,100]
plots = {
'hptplot': Hist(rptbins,None,'Highest Muon pT','Events passing HLT','upplots/TrigHpTplot'),
'ptplot': Hist(rptbins,None,'Highest Muon pT','Events','upplots/TrigpTplot'),
'ratioptplot': Hist(rptbins,None,'Highest Muon pT','HLT_Mu7_IP4 / Events with Muons of sip > 5','upplots/TrigRatiopTPlot'),
'sipplot': Hist(20,(0,20),'Highest Muon SIP', 'Events', 'upplots/TrigSIPplot'),
'hsipplot': Hist(20,(0,20),'Highest Muon SIP', 'Events', 'upplots/TrigHSIPplot'),
'ratiosipplot': Hist(20,(0,20),'Highest Muon SIP', 'HLT_Mu7_IP4 / Events with muons of pT > 10', 'upplots/TrigRatioSIPplot'),
'HLTcutflow': Hist(12,(-0.5,11.5),'All // HLT_Mu7/8/9/12_IP4/3,5,6/4,5,6/6','Events','upplots/cutflowHLT'),
'L1Tcutflow': Hist(12,(-0.5,11.5),'All // L1_SingleMu6/7/8/9/10/12/14/16/18','Events','upplots/cutflowL1T'),
'HLTcutflowL': Hist(12,(-0.5,11.5),'All // HLT_Mu7/8/9/12_IP4/3,5,6/4,5,6/6','Events','upplots/cutflowHLT-L'),
'L1TcutflowL': Hist(12,(-0.5,11.5),'All // L1_SingleMu6/7/8/9/10/12/14/16/18','Events','upplots/cutflowL1T-L')
}
cutflow2d = Hist2d([9,10],[[-0.5,8.5],[-0.5,9.5]],'All // HLT_Mu7/8/9/12_IP4/3,5,6/4,5,6/6',
'All // L1_SingleMu6/7/8/9/10/12/14/16/18','upplots/cutflowHLTvsL1T',files[0])
for plot in plots:
plots[plot].title = files[0]
## Create an internal figure for pyplot to write to
plt.figure(1)
## Loop over all input files
for fnum in range(len(files)):
print('Opening '+files[fnum])
## Open the file and retrieve our key branches
f = uproot.open(files[fnum])
events = f.get('Events')
HLTcuts = ['HLT_Mu7_IP4','HLT_Mu8_IP3','HLT_Mu8_IP5','HLT_Mu8_IP6','HLT_Mu9_IP4','HLT_Mu9_IP5','HLT_Mu9_IP6','HLT_Mu12_IP6']
L1Tcuts = ['L1_SingleMu6','L1_SingleMu7','L1_SingleMu8','L1_SingleMu9','L1_SingleMu10','L1_SingleMu12','L1_SingleMu14','L1_SingleMu16','L1_SingleMu18']
Muon = PhysObj('Muon',files[fnum],'pt','eta','phi','sip3d','mediumId')
Trig = PhysObj('trig')
HLT = PhysObj('HLTrig')
L1T = PhysObj('L1Trig')
Trig.vals = pd.DataFrame(events.array('HLT_Mu7_IP4_part0')).rename(columns=inc)
for tr in HLTcuts:
HLT[tr] = pd.DataFrame(events.array(tr+'_part0')).rename(columns=inc)
for tr in L1Tcuts:
L1T[tr]= pd.DataFrame(events.array(tr+'er1p5')).rename(columns=inc)
ev = Event(Muon,Trig,HLT,L1T)
print('Processing ' + str(len(Muon.pt)) + ' events')
## Fill 0 bin of cut flow plots
plots['HLTcutflow'].dfill(HLT[HLTcuts[0]]*0)
plots['L1Tcutflow'].dfill(L1T[L1Tcuts[0]]*0)
cutflow2d.dfill(HLT[HLTcuts[0]]*0,HLT[HLTcuts[0]]*0)
## Fill the rest of the bins
ct = 1
for i in HLT:
plots['HLTcutflow'].dfill(HLT[i][HLT[i]].dropna()*ct)
cutflow2d.dfill(HLT[i][HLT[i]].dropna()*ct,HLT[i][HLT[i]].dropna()*0)
ct = ct + 1
ct = 1
for i in L1T:
plots['L1Tcutflow'].dfill(L1T[i][L1T[i]].dropna()*ct)
cutflow2d.dfill(L1T[i][L1T[i]].dropna()*0,L1T[i][L1T[i]].dropna()*ct)
ct = ct + 1
ht = 1
for i in HLT:
lt = 1
for j in L1T:
cutflow2d.dfill(HLT[i][HLT[i] & L1T[j]].dropna()*ht,L1T[j][L1T[j] & HLT[i]].dropna()*lt)
lt = lt + 1
ht = ht + 1
##Perform global cuts
Muon.cut(abs(Muon.eta)<1.5)
Muon.cut(Muon.mediumId==True)
ev.sync()
##Fill bin 1 of cut flow lots
#plots['HLTcutflow'].fill((Muon.pt/Muon.pt).max(axis=1).dropna())
#plots['L1Tcutflow'].fill((Muon.pt/Muon.pt).max(axis=1).dropna())
## Cut muons and trim triggers to the new size
MuonP = Muon.cut(Muon.sip3d>5,split=True)
MuonS = Muon.cut(Muon.pt>10,split=True)
TrigP = Trig.trimTo(MuonP.pt,split=True)
TrigS = Trig.trimTo(MuonS.sip3d,split=True)
## Reshape triggers to fit our muons
for i in MuonP.pt.columns:
TrigP.vals[i] = TrigP.vals[1]
for i in MuonS.sip3d.columns:
TrigS.vals[i] = TrigS.vals[1]
## Create the two histograms we want to divide
plt.figure(1)
plots['ptplot'].fill(MuonP.pt.max(axis=1))
plots['hptplot'].fill(MuonP.pt[TrigP.vals].max(axis=1).dropna(how='all'))
plots['sipplot'].fill(MuonS.sip3d.max(axis=1))
plots['hsipplot'].fill(MuonS.sip3d[TrigS.vals].max(axis=1).dropna(how='all'))
plots['ratioptplot'].add(plots['hptplot'].divideby(plots['ptplot'],split=True))
plots['ratiosipplot'].add(plots['hsipplot'].divideby(plots['sipplot'],split=True))
plots['HLTcutflowL'].add(plots['HLTcutflow'])
plots['L1TcutflowL'].add(plots['L1Tcutflow'])
cutflow2d.norm()[0][0][0] = 0
cutflow2d.plot(text=True,edgecolor='black')
plots.pop('HLTcutflowL').norm().plot(ylim=(None,.2))
plots.pop('L1TcutflowL').norm().plot(ylim=(None,.2))
plots.pop('HLTcutflow').norm().plot()
plots.pop('L1Tcutflow').norm().plot()
for pl in plots:
plots[pl].plot()
sys.exit()
def ana(sigfiles,bgfiles,isLHE=False):
#%%################
# Plots and Setup #
###################
#training=True
#training=False
#tf.random.set_random_seed(2)
#tf.compat.v1.set_random_seed(2)
#np.random.seed(2)
#fig = plt.figure(figsize=(10.0,6.0))
global LOADMODEL
Skey = 'Signal'
Bkey = 'Background'
l1 = 8
l2 = 8
l3 = 8
alpha = 0.85
gamma = 0.8
# bmodel = keras.Sequential([
# keras.Input(shape=(4,),dtype='float32'),
# #keras.layers.Flatten(input_shape=(8,)),
# keras.layers.Dense(l1, activation=tf.nn.relu),
# keras.layers.Dense(l2, activation=tf.nn.relu),
# keras.layers.Dense(l3, activation=tf.nn.relu),
# #keras.layers.Dropout(0.1),
# keras.layers.Dense(1, activation=tf.nn.sigmoid),
# ])
# optimizer = keras.optimizers.Adam(learning_rate=0.01)
# bmodel.compile(optimizer=optimizer,#'adam',
# #loss='binary_crossentropy',
# #loss=[focal_loss],
# #loss=[custom],
# loss=[binary_focal_loss(alpha, gamma)],
# metrics=['accuracy'])#,tf.keras.metrics.AUC()])
#nbatch = math.floor(nbg / (2*nsig))
netvars = ['pt','eta','mass','CSVV2','DeepB','msoft','DDBvL','H4qvs']
bnetvars = ['DeepB','H4qvs','DDBvL','CSVV2']
pnetvars = ['pt','eta','mass','msoft']
## Define what pdgId we expect the A to have
Aid = 9000006
## How many resolved jets we want to target with our analysis
#resjets = 4
Aid = 36
## Make a dictionary of histogram objects
plots = {
"SigDist": Hist2d([20,20],[[0,1],[0,1]],'b-tag network','phys network','netplots/SigDist'),
"BGDist": Hist2d([20,20],[[0,1],[0,1]],'b-tag network','phys network','netplots/BGDist'),
"SigProfile": Hist(20,(0,1),'b-tag bin','phys network avg','netplots/SigProfile'),
"BGProfile": Hist(20,(0,1),'b-tag bin','phys network avg','netplots/BGProfile'),
"SigRanges": Hist(20,(0,1),'confidence','events','netplots/SigRanges'),
"BGRanges": Hist(20,(0,1),'confidence','events','netplots/BGRanges'),
}
for i in range(3):
plots.update({f"SigRanges{i}":cp.deepcopy(plots['SigRanges']),F"BGRanges{i}":cp.deepcopy(plots['BGRanges'])})
## Create an internal figure for pyplot to write to
plt.figure(1)
if isLHE:
nbg = len(bgfiles)/nlhe
if float(nbg).is_integer():
nbg = int(nbg)
else:
raise Exception('LHE argument specified, but BG files do not divide evenly into '+str(nlhe))
else:
nbg = len(bgfiles)
nsig = len(sigfiles)
sigmbg = nbg - nsig
## Loop over input files
for fnum in range(max(nbg,nsig)):
print('bg',nbg,'sig',nsig)
#####################
# Loading Variables #
#####################
if isLHE:
print('Opening',sigfiles[fnum],'+ LHE Background')
else:
print('Opening',sigfiles[fnum],'+',bgfiles[fnum])
## Loop some data if the bg/signal files need to be equalized
if sigmbg > 0:
print('Catching up signal')
sigfiles.append(sigfiles[fnum])
sigmbg = sigmbg - 1
elif sigmbg < 0:
if isLHE:
print('Catching up background')
for i in range(nlhe):
bgfiles.append(bgfiles[fnum+i])
sigmbg = sigmbg + 1
else:
print('Catching up background')
bgfiles.append(bgfiles[fnum])
sigmbg = sigmbg + 1
print('diff:',sigmbg)
## Open our file and grab the events tree
if isLHE:
bgevents = []
for i in range(nlhe):
idx = fnum*nlhe + i
print('Opening ',bgfiles[idx])
bgevents.append(uproot.open(bgfiles[idx]).get('Events'))
else:
bgf = uproot.open(bgfiles[fnum])
bgevents = bgf.get('Events')
sigf = uproot.open(sigfiles[fnum])
sigevents = sigf.get('Events')
def loadjets(jets, events,wname=''):
jets.eta= pd.DataFrame(events.array('FatJet_eta', executor=executor)).rename(columns=inc)
jets.phi= pd.DataFrame(events.array('FatJet_phi', executor=executor)).rename(columns=inc)
jets.pt = pd.DataFrame(events.array('FatJet_pt' , executor=executor)).rename(columns=inc)
jets.mass=pd.DataFrame(events.array('FatJet_mass', executor=executor)).rename(columns=inc)
jets.CSVV2 = pd.DataFrame(events.array('FatJet_btagCSVV2', executor=executor)).rename(columns=inc)
jets.DeepB = pd.DataFrame(events.array('FatJet_btagDeepB', executor=executor)).rename(columns=inc)
jets.DDBvL = pd.DataFrame(events.array('FatJet_btagDDBvL', executor=executor)).rename(columns=inc)
jets.msoft = pd.DataFrame(events.array('FatJet_msoftdrop', executor=executor)).rename(columns=inc)
jets.H4qvs = pd.DataFrame(events.array('FatJet_deepTagMD_H4qvsQCD', executor=executor)).rename(columns=inc)
jets.event = pd.DataFrame(events.array('event', executor=executor)).rename(columns=inc)
jets.npvs = pd.DataFrame(events.array('PV_npvs', executor=executor)).rename(columns=inc)
jets.npvsG = pd.DataFrame(events.array('PV_npvsGood', executor=executor)).rename(columns=inc)
jets.extweight = jets.event / jets.event
if wname != '':
weights = pickle.load(open('weights/'+wname+'-'+fstrip(DATANAME)+'.p',"rb" ))
for prop in ['genweights','PUweights','normweights']:
#print('jets.extweight[1]')#,jets.extweight[1])
jets.extweight[1] = jets.extweight[1] * weights[prop][1]
else:
jets.extweight = jets.event / jets.event
for j in range(1,jets.pt.shape[1]):
jets.event[j+1] = jets.event[1]
jets.npvs[j+1] = jets.npvs[1]
jets.npvsG[j+1] = jets.npvsG[1]
#if POSTWEIGHT:
jets.extweight[j+1] = jets.extweight[1]
return jets
sigjets = loadjets(PhysObj('sigjets'),sigevents,fstrip(sigfiles[fnum]))
if isLHE:
#bgjets = [PhysObj('300'),PhysObj('500'),PhysObj('700'),PhysObj('1000'),PhysObj('1500'),PhysObj('2000'),PhysObj('inf')]
bgjets = []
for i in range(nlhe):
bgjets.append(loadjets(PhysObj(str(i)),bgevents[i],fstrip(bgfiles[(fnum*nlhe)+i])))
else:
bgjets = loadjets(PhysObj('bgjets'),bgevents,fstrip(bgfiles[fnum]))
print(f"Processing {str(len(sigjets.eta))} {Skey} events")
if True:
pdgida = sigevents.array('GenPart_pdgId')
paridxa = sigevents.array('GenPart_genPartIdxMother')
parida = pdgida[paridxa]
bs = PhysObj('bs')
## Removes all particles that do not have A parents
## from the GenPart arrays, then removes all particles
## that are not bs after resizing the pdgid array to be a valid mask
bs.oeta = pd.DataFrame(sigevents.array('GenPart_eta')[abs(parida)==Aid][abs(pdgida)[abs(parida)==Aid]==5]).rename(columns=inc)
bs.ophi = pd.DataFrame(sigevents.array('GenPart_phi')[abs(parida)==Aid][abs(pdgida)[abs(parida)==Aid]==5]).rename(columns=inc)
bs.opt = pd.DataFrame(sigevents.array('GenPart_pt' )[abs(parida)==Aid][abs(pdgida)[abs(parida)==Aid]==5]).rename(columns=inc)
## Test b order corresponds to As
testbs = pd.DataFrame(sigevents.array('GenPart_genPartIdxMother')[abs(parida)==Aid][abs(pdgida)[abs(parida)==Aid]==5]).rename(columns=inc)
## The first term checks b4 has greater idx than b1, the last two check that the bs are paired
if ((testbs[4]-testbs[1]).min() <= 0) or ((abs(testbs[2]-testbs[1]) + abs(testbs[4])-testbs[3]).min() != 0):
print('b to A ordering violated - time to do it the hard way')
sys.exit()
As = PhysObj('As')
As.oeta = pd.DataFrame(sigevents.array('GenPart_eta', executor=executor)[abs(parida)==25][abs(pdgida)[abs(parida)==25]==Aid]).rename(columns=inc)
As.ophi = pd.DataFrame(sigevents.array('GenPart_phi', executor=executor)[abs(parida)==25][abs(pdgida)[abs(parida)==25]==Aid]).rename(columns=inc)
As.opt = pd.DataFrame(sigevents.array('GenPart_pt' , executor=executor)[abs(parida)==25][abs(pdgida)[abs(parida)==25]==Aid]).rename(columns=inc)
As.omass =pd.DataFrame(sigevents.array('GenPart_mass', executor=executor)[abs(parida)==25][abs(pdgida)[abs(parida)==25]==Aid]).rename(columns=inc)
higgs = PhysObj('higgs')
higgs.eta = pd.DataFrame(sigevents.array('GenPart_eta', executor=executor)[abs(parida)!=25][abs(pdgida)[abs(parida)!=25]==25]).rename(columns=inc)
higgs.phi = pd.DataFrame(sigevents.array('GenPart_phi', executor=executor)[abs(parida)!=25][abs(pdgida)[abs(parida)!=25]==25]).rename(columns=inc)
higgs.pt = pd.DataFrame(sigevents.array('GenPart_pt' , executor=executor)[abs(parida)!=25][abs(pdgida)[abs(parida)!=25]==25]).rename(columns=inc)
slimjets = PhysObj('slimjets')
slimjets.eta= pd.DataFrame(sigevents.array('Jet_eta', executor=executor)).rename(columns=inc)
slimjets.phi= pd.DataFrame(sigevents.array('Jet_phi', executor=executor)).rename(columns=inc)
slimjets.pt = pd.DataFrame(sigevents.array('Jet_pt' , executor=executor)).rename(columns=inc)
slimjets.mass=pd.DataFrame(sigevents.array('Jet_mass', executor=executor)).rename(columns=inc)
#sigjets.CSVV2 = pd.DataFrame(sigevents.array('FatJet_btagCSVV2')).rename(columns=inc)
slimjets.DeepB = pd.DataFrame(sigevents.array('Jet_btagDeepB', executor=executor)).rename(columns=inc)
#sigjets.DDBvL = pd.DataFrame(sigevents.array('FatJet_btagDDBvL')).rename(columns=inc)
#sigjets.msoft = pd.DataFrame(sigevents.array('FatJet_msoftdrop')).rename(columns=inc)
slimjets.DeepFB= pd.DataFrame(sigevents.array('Jet_btagDeepFlavB', executor=executor)).rename(columns=inc)
slimjets.puid = pd.DataFrame(sigevents.array('Jet_puId', executor=executor)).rename(columns=inc)
## Figure out how many bs and jets there are
nb = bs.oeta.shape[1]
njet= sigjets.eta.shape[1]
#nsjet=slimjets.eta.shape[1]
na = As.oeta.shape[1]
if na != 2:
print("More than two As per event, found "+str(na)+", halting")
sys.exit()
## Create sorted versions of A values by pt
for prop in ['eta','phi','pt','mass']:
As[prop] = pd.DataFrame()
for i in range(1,3):
As[prop][i] = As['o'+prop][As.opt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
## Clean up original ordered dataframes; we don't really need them
#del As['o'+prop]
## Reorder out b dataframes to match sorted A parents
tframe = pd.DataFrame()
tframe[1] = (As.opt.rank(axis=1,ascending=False,method='first')==1)[1]
tframe[2] = (As.opt.rank(axis=1,ascending=False,method='first')==1)[1]
tframe[3] = (As.opt.rank(axis=1,ascending=False,method='first')==1)[2]
tframe[4] = (As.opt.rank(axis=1,ascending=False,method='first')==1)[2]
for prop in ['eta','phi','pt']:
bs[prop] = pd.DataFrame()
bs[prop][1] = bs['o'+prop][tframe][1].dropna().append(bs['o'+prop][tframe][3].dropna()).sort_index()
bs[prop][2] = bs['o'+prop][tframe][2].dropna().append(bs['o'+prop][tframe][4].dropna()).sort_index()
bs[prop][3] = bs['o'+prop][~tframe][1].dropna().append(bs['o'+prop][~tframe][3].dropna()).sort_index()
bs[prop][4] = bs['o'+prop][~tframe][2].dropna().append(bs['o'+prop][~tframe][4].dropna()).sort_index()
## Clean up original ordered dataframes; we don't really need them.
#del bs['o'+prop]
# ## Sort our b dataframes in descending order of pt
# for prop in ['spt','seta','sphi']:
# bs[prop] = pd.DataFrame()
# #bs.spt, bs.seta, bs.sphi = pd.DataFrame(), pd.DataFrame(), pd.DataFrame()
# for i in range(1,nb+1):
# bs[prop][i] = bs[prop[1:]][bs.pt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
# #bs.seta[i] = bs.eta[bs.pt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
# #bs.sphi[i] = bs.phi[bs.pt.rank(axis=1,ascending=False,method='first')==i].max(axis=1)
# plots['genAmass'].dfill(As.mass)
ev = Event(bs,sigjets,As,higgs)
if isLHE:
for jets in bgjets+[sigjets]:
jets.cut(jets.pt > 170)#240)#170)
jets.cut(abs(jets.eta)<2.4)
jets.cut(jets.DDBvL > 0.8)#0.8)#0.6)
jets.cut(jets.DeepB > 0.4184)
jets.cut(jets.msoft > 90)#90)#0.25)
#
jets.cut(jets.mass > 90)
else:
for jets in [bgjets, sigjets]:
jets.cut(jets.pt > 170)#170)
jets.cut(abs(jets.eta)<2.4)
jets.cut(jets.DDBvL > 0.8)#0.6)
jets.cut(jets.DeepB > 0.4184)
jets.cut(jets.msoft > 90)#0.25)
#
jets.cut(jets.mass > 90)
bs.cut(bs.pt>5)
bs.cut(abs(bs.eta)<2.4)
ev.sync()
slimjets.cut(slimjets.DeepB > 0.1241)
slimjets.cut(slimjets.DeepFB > 0.277)
slimjets.cut(slimjets.puid > 0)
slimjets.trimto(jets.eta)
##############################
# Processing and Calculation #
##############################
if True:
## Create our dR dataframe by populating its first column and naming it accordingly
jbdr2 = pd.DataFrame(np.power(sigjets.eta[1]-bs.eta[1],2) + np.power(sigjets.phi[1]-bs.phi[1],2)).rename(columns={1:'Jet 1 b 1'})
sjbdr2= pd.DataFrame(np.power(slimjets.eta[1]-bs.eta[1],2) + np.power(slimjets.phi[1]-bs.phi[1],2)).rename(columns={1:'Jet 1 b 1'})
## Loop over jet x b combinations
jbstr = []
for j in range(1,njet+1):
for b in range(1,nb+1):
## Make our column name
jbstr.append("Jet "+str(j)+" b "+str(b))
if (j+b==2):
continue
## Compute and store the dr of the given b and jet for every event at once
jbdr2[jbstr[-1]] = pd.DataFrame(np.power(sigjets.eta[j]-bs.eta[b],2) + np.power(sigjets.phi[j]-bs.phi[b],2))
sjbdr2[jbstr[-1]]= pd.DataFrame(np.power(slimjets.eta[j]-bs.eta[b],2) + np.power(slimjets.phi[j]-bs.phi[b],2))
## Create a copy array to collapse in jets instead of bs
blist = []
sblist = []
for b in range(nb):
blist.append(np.sqrt(jbdr2.filter(like='b '+str(b+1))))
blist[b] = blist[b][blist[b].rank(axis=1,method='first') == 1]
blist[b] = blist[b].rename(columns=lambda x:int(x[4:6]))
sblist.append(np.sqrt(sjbdr2.filter(like='b '+str(b+1))))
sblist[b] = sblist[b][sblist[b].rank(axis=1,method='first') == 1]
sblist[b] = sblist[b].rename(columns=lambda x:int(x[4:6]))
## Trim resolved jet objects
# if resjets==3:
# for i in range(nb):
# for j in range(nb):
# if i != j:
# blist[i] = blist[i][np.logical_not(blist[i] > blist[j])]
# blist[i] = blist[i][blist[i]<0.4]
## Cut our events to only events with 3-4 bs in one fatjet of dR<0.8
fjets = blist[0][blist[0]<0.8].fillna(0)/blist[0][blist[0]<0.8].fillna(0)
for i in range(1,4):
fjets = fjets + blist[i][blist[i]<0.8].fillna(0)/blist[i][blist[i]<0.8].fillna(0)
fjets = fjets.max(axis=1)
# fjetsfail = fjets[fjets!=4].dropna()
fjets = fjets[fjets==4].dropna()
# ev.sync()
# sigjetsfail = sigjets.trimto(fjetsfail,split=True)
sigjets.trimto(fjets)
ev.sync()
##################################
# Preparing Neural Net Variables #
##################################
bgjetframe = pd.DataFrame()
extvars = ['event','extweight','npvs','npvsG']
if isLHE:
bgpieces = []
wtpieces = []
for i in range(nlhe):
tempframe = pd.DataFrame()
twgtframe = pd.DataFrame()
for prop in netvars+extvars:
twgtframe[prop] = bgjets[i][prop][bgjets[i]['pt'].rank(axis=1,method='first') == 1].max(axis=1)
if 'eta' in netvars:
twgtframe['eta'] = abs(twgtframe['eta'])
twgtframe['val'] = 0
tempframe = twgtframe.sample(frac=lheweights[i],random_state=6)
twgtframe['extweight'] = twgtframe['extweight'] * lheweights[i]
bgpieces.append(tempframe)
#pickle.dump(tempframe, open(filefix+str(i)+"piece.p", "wb"))
wtpieces.append(twgtframe)
bgjetframe = pd.concat(bgpieces,ignore_index=True)
bgrawframe = pd.concat(wtpieces,ignore_index=True)
bgjetframe = bgjetframe.dropna()
bgrawframe = bgrawframe.dropna()
bgtrnframe = bgjetframe[bgjetframe['event']%2 == 0]
else:
for prop in netvars + extvars:
bgjetframe[prop] = bgjets[prop][bgjets['pt'].rank(axis=1,method='first') == 1].max(axis=1)
bgjetframe['eta'] = abs(bgjetframe['eta'])
bgjetframe['val'] = 0
bgtrnframe = bgjetframe[bgjetframe['event']%2 == 0]
nbg = bgtrnframe.shape[0]
sigjetframe = pd.DataFrame()
for prop in netvars + extvars:
sigjetframe[prop] = sigjets[prop][sigjets['pt'].rank(axis=1,method='first') == 1].max(axis=1)
if 'eta' in netvars:
sigjetframe['eta'] = abs(sigjetframe['eta'])
sigjetframe['val'] = 1
sigtrnframe = sigjetframe[sigjetframe['event']%2 == 0]
nsig = sigtrnframe.shape[0]
# sigjetfailframe = pd.DataFrame()
# for prop in netvars + extvars:
# sigjetfailframe[prop] = sigjetsfail[prop][sigjetsfail['pt'].rank(axis=1,method='first') == 1].max(axis=1)
# if 'eta' in netvars:
# sigjetfailframe['eta'] = abs(sigjetfailframe['eta'])
# sigjetfailframe['val'] = 1
# sigtrnFframe = sigjetfailframe[sigjetfailframe['event']%2 == 0]
# nsig = sigtrnFframe.shape[0]
print(f"{Skey} cut to {sigjetframe.shape[0]} events")
print(f"{Bkey} has {bgjetframe.shape[0]} intended events")
extvars = extvars + ['val']
#######################
# Training Neural Net #
#######################
if isLHE:
bgjetframe=bgrawframe
X_inputs = pd.concat([bgjetframe,sigjetframe])
# XF_inputs = pd.concat([bgjetframe,sigjetfailframe])
W_inputs = X_inputs['extweight']
# WF_inputs = XF_inputs['extweight']
Y_inputs = X_inputs['val']
# YF_inputs = XF_inputs['val']
X_inputs = X_inputs.drop(extvars,axis=1)
# XF_inputs = XF_inputs.drop(extvars,axis=1)
bmodel = keras.models.load_model('btagfiles/weighted.hdf5', compile=False)
physmodel = keras.models.load_model('physfiles/weighted.hdf5', compile=False)
bscaler = pickle.load( open("btagfiles/weightedscaler.p", "rb" ) )
physcaler = pickle.load( open("physfiles/weightedscaler.p", "rb" ) )
##
#print(scaler.transform(bgpieces[1].drop('val',axis=1)))
##
Xb_inputs = bscaler.transform(X_inputs.drop(pnetvars,axis=1))
# XbF_inputs = bscaler.transform(XF_inputs)
Xp_inputs = physcaler.transform(X_inputs.drop(bnetvars,axis=1))
# XpF_inputs = physcaler.transform(XF_inputs)
##################################
# Analyzing and Plotting Outputs #
##################################
distsb = bmodel.predict( Xb_inputs[Y_inputs==1])
distsp = physmodel.predict(Xp_inputs[Y_inputs==1])
distbb = bmodel.predict (Xb_inputs [Y_inputs==0])
distbp = physmodel.predict (Xp_inputs [Y_inputs==0])
plots['SigDist' ].fill(distsb[:,0] ,distsp[:,0] ,weights=W_inputs[Y_inputs==1])
plots['BGDist' ].fill(distbb[:,0] ,distbp[:,0] ,weights=W_inputs[Y_inputs==0])
Sprofile, Serr, Bprofile, Berr = [],[],[],[]
for i in range(plots['SigDist'][0].shape[0]):
Sprofile.append(np.average(distsp[np.logical_and(
distsb > plots['SigDist'][1][i],
distsb <= plots['SigDist'][1][i+1])]))
Serr.append(np.std(distsp[np.logical_and(
distsb > plots['SigDist'][1][i],
distsb <= plots['SigDist'][1][i+1])]))
Bprofile.append(np.average(distbp[np.logical_and(
distbb > plots['BGDist'][1][i],
distbb <= plots['BGDist'][1][i+1])]))
Berr.append(np.std(distbp[np.logical_and(
distbb > plots['BGDist'][1][i],
distbb <= plots['BGDist'][1][i+1])]))
plots['SigProfile'][0] = Sprofile
plots['SigProfile'].ser = np.power(Serr,2)
plots['BGProfile'][0] = Bprofile
plots['BGProfile'].ser = np.power(Berr,2)
sbranges, bbranges = [0.0], [0.0]
for i in range(1,4):
sbranges.append(np.sort(distsb.ravel())[math.floor(distsb.shape[0]*i/3)-1])
bbranges.append(np.sort(distbb.ravel())[math.floor(distbb.shape[0]*i/3)-1])
for i in range(3):
plots[f"SigRanges{i}"].fill(distsp[np.logical_and(
distsb > sbranges[i],
distsb <= sbranges[i+1])])
plots[f"BGRanges{i}"].fill(distbp[np.logical_and(
distbb > bbranges[i],
distbb <= bbranges[i+1])])
for p in [plots['SigDist'],plots['BGDist']]:
p.plot()
for p in [plots['SigProfile'],plots['BGProfile']]:
p.plot(error=True,htype='err')
plt.clf()
for plot in ['SigRanges','BGRanges']:
plots[f"{plot}{0}"].make(linestyle='-',color='b',htype='step')
plots[f"{plot}{1}"].make(linestyle='--',color='r',htype='step')
plots[f"{plot}{2}"].make(linestyle=':',color='k',htype='step')
plots[plot].plot(same=True,legend=['Low Confidence','Medium Confidence','High Confidence'])