channel_ = d["channel"] y2 = 0 inputs = [] if(int(d["channel"]) == 1): for var in mvavars1: inputs.append(float(d[var])) y2 = bnn_4mu_loosesel4_ext.bnn_4mu_loosesel4(inputs, 100, 200); if(int(d["channel"]) == 2): for var in mvavars2: inputs.append(float(d[var])) y2 = bnn_4e_loosesel4_ext.bnn_4e_loosesel4(inputs, 100, 200); if(int(d["channel"]) == 3): for var in mvavars3: inputs.append(float(d[var])) y2 = bnn_2e2mu_loosesel4_ext.bnn_2e2mu_loosesel4(inputs, 100, 200); if(applycuts(columns2, data_, int(d["channel"]))): btotal_sel += (weight) else: htempps.Fill(y, y2) i+=1 if( not (i%10000)): print i print y hs.Draw() c1.Update() sig.close() print "Integral hs: " + str(hs.Integral())
def eventloop(sig,bkg,channel,evtset):
line = sig.readline()
data = line.rsplit()
columns = []
for varname in data:
print varname
varname = varname.lstrip(':b')
columns.append(varname);
bins = 200
heff = TH2F('heff_train', 'heff_train', bins, 0, 1.0,bins, 0.0, 1)
heffmela = TH2F('heff_mela', 'heff_mela', bins, 0, 1.0,bins, 0.0, 1)
hbnnvsmela_zz = TH2F('hbnnvsmela', 'hbnnvsmela', 20, 0, 1, 20, 0, 1)
hbnnvsmela_sig = TH2F('hbnnvsmela', 'hbnnvsmela', 20, 0, 1, 20, 0, 1)
hbnnvsmela_zx = TH2F('hbnnvsmela', 'hbnnvsmela', 20, 0, 1, 20, 0, 1)
hs = TH1F('mva_s', 'mva_s', bins, 0, 1)
hs.SetLineColor(2);
hb = TH1F('mva_b', 'mva_b', bins, 0, 1)
hb.SetLineColor(1);
hs.Draw()
hb.Draw("same")
hb_zx = TH1F('hb_zx', 'hb_zx', 20, 0, 1)
hb_zx.SetLineColor(2);
hb_zx.Draw()
hb_zx_rw = TH1F('hb_zx', 'hb_zx', 20, 0, 1)
hb_zx_rw.SetLineColor(2);
hb_zx_rw.Draw()
hb_zx_sel = TH1F('hb_zx', 'hb_zx', 20, 0, 1)
hb_zx_sel.SetLineColor(2);
hb_zx_sel.Draw()
hmelas = TH1F('mela_s', 'mela_s', bins, 0, 1)
hmelas.SetLineColor(2);
hmelab = TH1F('mela_b', 'mela_b', bins, 0, 1)
hmelab.SetLineColor(1);
hmelas.Draw()
hmelab.Draw("same")
i = 0
c1 = TCanvas( 'c1', 'mva output', 200, 10, 700, 500 )
c1.cd()
c1.SetGridx()
c1.SetGridy()
c1.Draw()
gStyle.SetOptStat(0);
#------- sig events ---------
stotal_ = 0
stotal_sel = 0
sigweight = 0
#do signal
for line in sig:
data = line.rsplit()
# data_ = [float(x) for x in line.rsplit()]
data_ = map(float, line.rsplit())
d = dict(zip(columns, data))
inputs = []
if(float(d["mass4l"]) < 110):
continue
if(float(d["mass4l"]) > 150):
continue
channel_ = int(d["channel"])
weight = float(d["weight"])
if(int(d["channel"]) == 1):
for var in mvavars1:
inputs.append(float(d[var]))
y = bnn_4mu_loosesel4_ext.bnn_4mu_loosesel4(inputs, 100, 200);
if(int(d["channel"]) == 2):
for var in mvavars2:
inputs.append(float(d[var]))
y = bnn_4e_loosesel4_ext.bnn_4e_loosesel4(inputs, 100, 200);
if(int(d["channel"]) == 3):
for var in mvavars3:
inputs.append(float(d[var]))
y = bnn_2e2mu_loosesel4_ext.bnn_2e2mu_loosesel4(inputs, 100, 200);
weight = float(d["weight"])
stotal_ += float(d["weight"])
if(applycuts(columns, data_, channel_)):
stotal_sel += float(d["weight"])
# continue
else:
hmelas.Fill(float(d["mela"]), weight)
hs.Fill(y, weight)
hbnnvsmela_sig.Fill(y, float(d["mela"]), weight)
if(int(d["index"]) == 6):
sigweight+=weight
i+=1
if( not (i%10000)):
print i
print y
hs.Draw()
c1.Update()
sig.close()
print "Higgs 125 weight: " + str(sigweight)
print "Integral hs: " + str(hs.Integral())
hs.Scale(1/hs.Integral())
hmelas.Scale(1/hmelas.Integral())
#do background
line = bkg.readline()
data = line.rsplit()
columns = []
for varname in data:
print varname
varname = varname.lstrip(':b')
columns.append(varname);
#------- ZZ events ---------
average = 0
btotal_ = 0
btotal_sel = 0
i = 0
for line in bkg:
data = line.rsplit()
# data_ = [float(x) for x in line.rsplit()]
data_ = map(float, line.rsplit())
d = dict(zip(columns, data))
channel_ = d["channel"]
if(float(d["mass4l"]) < 110):
continue
if(float(d["mass4l"]) > 150):
continue
weight = float(d["weight"])
y = 0
inputs = []
if(int(d["channel"]) == 1):
for var in mvavars1:
inputs.append(float(d[var]))
y = bnn_4mu_loosesel4_ext.bnn_4mu_loosesel4(inputs, 100, 200);
if(int(d["channel"]) == 2):
for var in mvavars2:
inputs.append(float(d[var]))
y = bnn_4e_loosesel4_ext.bnn_4e_loosesel4(inputs, 100, 200);
if(int(d["channel"]) == 3):
for var in mvavars3:
inputs.append(float(d[var]))
y = bnn_2e2mu_loosesel4_ext.bnn_2e2mu_loosesel4(inputs, 100, 200);
if(d["mela"] == "-nan"):
d["mela"]= "0"
btotal_ += float(weight)
if(applycuts(columns, data_, int(d["channel"]))):
# print "notselected"
btotal_sel += (weight)
# continue
else:
hmelab.Fill(float(d["mela"]), weight)
hb.Fill(y, weight)
hbnnvsmela_zz.Fill(y, float(d["mela"]), weight)
i+=1
if( not (i%10000)):
print i
print y
c1.Update();
average += float(d["mass4l"])
# print "average: " + str(average/i)
print "btotal_: " + str(btotal_)
print "Integral hb: " + str(hb.Integral())
dat = open('/home/jbochenek/data/HZZ4l_2012_data/hzz_loosesel/zx_mela.dat')
line = dat.readline()
data = line.rsplit()
columns = []
for varname in data:
print varname
varname = varname.lstrip(':b')
columns.append(varname);
#------- Z+X events ---------
f_fake = TFile('../fake_reweight.root')
f_fake.cd();
cratio_Run2012_e=gDirectory.Get("cratio_Run2012_e");
cratio_Run2012_mu= gDirectory.Get("cratio_Run2012_mu");
for line in dat:
data = line.rsplit()
data_ = map(float, line.rsplit())
d = dict(zip(columns, data))
reweight = 1
inputs = []
inputs = []
if(int(d["channel"]) == 1):
for var in mvavars1:
inputs.append(float(d[var]))
y = bnn_4mu_loosesel4_ext.bnn_4mu_loosesel4(inputs, 100, 200);
reweight *= cratio_Run2012_mu.Interpolate(float(d["lept3_eta"]), float(d["lept3_pt"])) * cratio_Run2012_mu.Interpolate(float(d["lept4_eta"]), float(d["lept4_pt"]));
if(int(d["channel"]) == 2):
for var in mvavars2:
inputs.append(float(d[var]))
y = bnn_4e_loosesel4_ext.bnn_4e_loosesel4(inputs, 100, 200);
reweight *= cratio_Run2012_e.Interpolate(float(d["lept3_eta"]), float(d["lept3_pt"])) * cratio_Run2012_e.Interpolate(float(d["lept4_eta"]), float(d["lept4_pt"]));
if(int(d["channel"]) == 3):
for var in mvavars3:
inputs.append(float(d[var]))
y = bnn_2e2mu_loosesel4_ext.bnn_2e2mu_loosesel4(inputs, 100, 200);
reweight *= cratio_Run2012_e.Interpolate(float(d["lept3_eta"]), float(d["lept3_pt"])) * cratio_Run2012_e.Interpolate(float(d["lept4_eta"]), float(d["lept4_pt"]));
weight = float(d["weight"])
d["issamesign"] = "0"
if(d["mela"] == "-nan"):
d["mela"] = "0"
btotal_ += float(d["weight"])
d["lept3_pfx"] = "0"
d["lept4_pfx"] = "0"
d["lept3_mvaid"] = "1.0"
d["lept4_mvaid"] = "1.0"
for k, v in enumerate(data_):
data_[k] = float(d[columns[k]])
# print columns[k] + " " + str(v)
if(applycuts(columns, data_, int(d["channel"]))):
# print "notselected"
btotal_sel += weight
else:
# hb_zx_sel.Fill(float(d["mela"]), 1.)
hmelab.Fill(float(d["mela"]), reweight)
hbnnvsmela_zx.Fill(y, float(d["mela"]), weight)
hb.Fill(y, reweight)
hb_zx_rw.Fill(float(d["mela"]), reweight)
hb_zx.Fill( float(d["mela"]), 1.)
i+=1
if( not (i%10000)):
print i
print y
hb.Draw("same")
c1.Update();
average += float(d["mass4l"])
hmelas.Fill(0.0, stotal_sel)
hmelab.Fill(0.0, btotal_sel)
# hs.Fill(0.0, stotal_sel)
# hb.Fill(0.0, btotal_sel)
# print "average: " + str(average/i)
print "btotal_: " + str(btotal_)
print "Integral hb: " + str(hb.Integral())
print "Integral zx: " + str(hb_zx.Integral())
print "Integral zx (sel.): " + str(hb_zx_sel.Integral())
print "Integral zx (rw): " + str(hb_zx_rw.Integral())
#------- Print Z+X plots ---------
c_zx = TCanvas( 'c_zx', 'mva zx output', 200, 10, 700, 500 )
c_zx.cd()
c_zx.SetGridx()
c_zx.SetGridy()
c_zx.Draw()
gStyle.SetOptStat(0);
hb_zx.Scale(1/hb_zx.Integral())
# hb_zx_sel.Scale(1/hb_zx_sel.Integral())
hb_zx_rw.Scale(1/hb_zx_rw.Integral())
c_zx.cd()
# c_zx.SetLogy()
hb_zx.SetTitle("BNN Discriminant: Z+X");
hb_zx.GetXaxis().SetTitle("D(x)");
legend = TLegend(0.3, 0.72, 0.68, 0.88)
hb_zx.SetLineColor(3)
hb_zx.Draw()
hb_zx_rw.SetLineColor(2)
hb_zx_rw.SetLineWidth(2)
hb_zx.SetLineWidth(2)
hb_zx_sel.SetLineWidth(2)
hb_zx_rw.Draw("same")
hb_zx_sel.SetLineColor(1)
# hb_zx_sel.Draw("same")
hb_zx.Draw("hist same")
legend.AddEntry(hb_zx_rw, "reweight", "l");
legend.AddEntry(hb_zx, "unweighted", "l");
# legend.AddEntry(hb_zx_sel, "selected", "l");
legend.Draw("same")
c_zx.Update()
c_zx.SaveAs("zx-reweight_" + evtset + ".gif")
#------- Print 2D plots ---------
c_2d = TCanvas( 'c_2d', 'mva zx output', 200, 10, 500, 500 )
c_2d.cd()
c_2d.SetGridx()
c_2d.SetGridy()
c_2d.Draw()
gStyle.SetOptStat(0);
c_2d.cd()
hbnnvsmela_zz.SetTitle("BNN/MELA");
hbnnvsmela_zz.GetXaxis().SetTitle("BNN(x)");
hbnnvsmela_zz.GetYaxis().SetTitle("MELA(x)");
legend = TLegend(0.22, 0.75, 0.78, 0.88)
hbnnvsmela_zx.SetMarkerColor(3)
# hbnnvsmela_zx.Draw()
hbnnvsmela_zz.SetMarkerColor(2)
hbnnvsmela_sig.SetMarkerColor(4)
hbnnvsmela_zz.SetFillColor(2)
hbnnvsmela_sig.SetFillColor(4)
hbnnvsmela_zx.SetFillColor(3)
hbnnvsmela_zz.SetLineWidth(2)
hbnnvsmela_zx.SetLineWidth(2)
hbnnvsmela_sig.SetLineWidth(2)
hbnnvsmela_zz.SetMarkerSize(2.);
hbnnvsmela_sig.SetMarkerSize(2.);
hbnnvsmela_zz.Draw("box")
# hbnnvsmela_sig.SetMarkerColor(1)
hbnnvsmela_sig.Draw("same box")
legend.AddEntry(hbnnvsmela_zz, "ZZ: corr. " + str(round(hbnnvsmela_zz.GetCorrelationFactor(), 3)), "p");
# legend.AddEntry(hbnnvsmela_zx, "Z+X", "p");
legend.AddEntry(hbnnvsmela_sig, "sig.: corr. " + str(round(hbnnvsmela_sig.GetCorrelationFactor(), 3)), "p");
legend.Draw("same")
c_2d.Update()
c_2d.SaveAs("bnn_vs_mela_" + evtset + ".gif")
line = dat.readline()
data = line.rsplit()
columns = []
for varname in data:
print varname
varname = varname.lstrip(':b')
columns.append(varname);
print "btotal_: " + str(btotal_)
print "Integral hb: " + str(hb.Integral())
bkg.close()
#------- BNN eff ---------
geff = TGraph()
geff.SetName("bnneff")
gmelaeff = TGraph()
gmelaeff.SetName("melaeff")
stotal=hs.Integral();
btotal=hb.Integral();
cs = cdf(hs);
cb = cdf(hb);
maxsig = 0;
optmvax = 0;
optmvay = 0;
for i in range(hs.GetNbinsX()):
cs1 = stotal - cs[i-1]
cb1 = btotal - cb[i-1]
es = cs1 / stotal
eb = cb1 / btotal
sign=0
if(cb1 > 0):
sign = cs1/sqrt(cs1 + cb1)
if(sign > maxsig):
maxsig = sign;
optcut = i;
optmvax = eb;
optmvay = es;
heff.Fill(eb, es);
if(i<2):
continue
else:
geff.SetPoint(geff.GetN(), eb, es);
print "\n Maximum Significance " + str(maxsig) + ", is bin: " + str(optcut) + ", which is BNN Cut: " + str(optcut/bins)
print "es: " + str(optmvay) + "\teb: " + str(optmvax)
#------- MELA eff ---------
smelatotal=hmelas.Integral()
bmelatotal=hmelab.Integral()
cmelas = cdf(hmelas);
cmelab = cdf(hmelab);
maxsig = 0;
optmvax = 0;
optmvay = 0;
for i in range(hmelas.GetNbinsX()):
cs1 = smelatotal - (cmelas[i-1])
cb1 = bmelatotal - (cmelab[i-1])
es = cs1 / smelatotal
eb = cb1 / bmelatotal
sign=0
if(cb1 > 0):
sign = cs1/sqrt(cs1 + cb1)
if(sign > maxsig):
maxsig = sign;
optcut = i;
optmvax = eb;
optmvay = es;
heffmela.Fill(eb, es);
if(i < 2):
continue
else:
gmelaeff.SetPoint(gmelaeff.GetN(), eb, es);
# print str(i) + " eb: " + str(eb) + " es: " + str(es)
print "\n Maximum Significance " + str(maxsig) + ", is bin: " + str(optcut) + ", which is MELA Cut: " + str(optcut/bins)
print "es: " + str(optmvay) + "\teb: " + str(optmvax)
print "stotal: " + str(stotal_)
print "btotal: " + str(btotal_)
print "stotal_sel: " + str(stotal_sel)
print "btotal_sel: " + str(btotal_sel)
ceffmela = TCanvas( 'ceffmela', 'effmela', 200, 10, 700, 500 )
ceffmela.cd();
gStyle.SetOptStat(0);
heffmela.SetMarkerSize(2);
ceffmela.Update();
ceffmela.SaveAs("heffmela_" + evtset + ".gif")
ceffmela.Update();
ceff = TCanvas( 'ceff', 'eff', 200, 10, 700, 500 )
ceff.cd();
legend_x = TLegend(0.6, 0.25, 0.85, 0.45)
gStyle.SetOptStat(0);
heff.SetLineColor(2);
geff.SetLineColor(2);
geff.SetLineWidth(2);
heff.SetTitle("ROC for BNN & MELA (bkg = ZZ & Z+X)");
heff.GetXaxis().SetTitle("#epsilon_{bkg.}");
heff.GetXaxis().SetTitleSize(0.06);
heff.GetYaxis().SetTitleSize(0.06);
heff.GetXaxis().SetTitleOffset(0.65);
heff.GetYaxis().SetTitleOffset(0.65);
heff.GetYaxis().SetTitle("#epsilon_{sig.}");
heff.Draw();
geff.Draw("same");
heffmela.SetLineColor(3);
gmelaeff.SetLineColor(3);
gmelaeff.SetLineWidth(2);
heffmela.Draw("same");
gmelaeff.Draw("same")
legend_x.AddEntry(heffmela, "MELA", "l");
legend_x.AddEntry(heff, "BNN", "l");
legend_x.Draw("same")
ceff.Update();
ceff.SaveAs("heff_" + evtset + ".gif")
ceff.Update();
#------- BNN Dist ---------
# c1.SetLogy();
hb.Scale(1/hb.Integral())
hmelab.Scale(1/hmelab.Integral())
c1.cd()
gStyle.SetOptStat(0);
hs.Rebin(2);
hb.Rebin(2);
hs.SetTitle("BNN Discriminant: sig. & bkg (ZZ & Z+X)");
hs.GetXaxis().SetTitle("D(x)");
hs.SetLineWidth(2)
hb.SetLineWidth(2)
hs.GetXaxis().SetTitleSize(0.06);
hs.GetYaxis().SetTitleSize(0.06);
hs.GetXaxis().SetTitleOffset(0.65);
hs.GetYaxis().SetTitleOffset(0.65);
hs.Draw()
hb.Draw("same")
c1.Update()
c1.SaveAs("hdist_bnn_" + evtset + ".gif")
#------- MELA Dist ---------
c2 = TCanvas( 'c2', 'mva output', 200, 10, 700, 500 )
c2.cd()
gStyle.SetOptStat(0);
c2.SetGridx()
c2.SetGridy()
c2.Draw()
# c2.SetLogy();
# hmelas.Scale(1/hmelas.Integral())
# hmelab.Scale(1/hmelab.Integral())
gStyle.SetOptStat(0);
hmelas.Rebin(2);
hmelab.Rebin(2);
hmelab.SetTitle("MELA Discriminant: sig. & bkg (ZZ & Z+X)");
hmelab.GetXaxis().SetTitle("D(x)");
hmelab.SetLineWidth(2)
hmelas.SetLineWidth(2)
hmelab.GetXaxis().SetTitleSize(0.06);
hmelab.GetYaxis().SetTitleSize(0.06);
hmelab.GetXaxis().SetTitleOffset(0.65);
hmelab.GetYaxis().SetTitleOffset(0.65);
hmelab.Draw()
hmelas.Draw("same")
c2.Update()
c2.SaveAs("hdist_mela_" + evtset + ".gif")
def eventloop(sig, bkg, channel, evtset):
line = sig.readline()
data = line.rsplit()
columns = []
for varname in data:
print varname
varname = varname.lstrip(':b')
columns.append(varname)
bins = 200
heff = TH2F('heff_train', 'heff_train', bins, 0, 1.0, bins, 0.0, 1)
heffmela = TH2F('heff_mela', 'heff_mela', bins, 0, 1.0, bins, 0.0, 1)
hbnnvsmela_zz = TH2F('hbnnvsmela', 'hbnnvsmela', 20, 0, 1, 20, 0, 1)
hbnnvsmela_sig = TH2F('hbnnvsmela', 'hbnnvsmela', 20, 0, 1, 20, 0, 1)
hbnnvsmela_zx = TH2F('hbnnvsmela', 'hbnnvsmela', 20, 0, 1, 20, 0, 1)
hs = TH1F('mva_s', 'mva_s', bins, 0, 1)
hs.SetLineColor(2)
hb = TH1F('mva_b', 'mva_b', bins, 0, 1)
hb.SetLineColor(1)
hs.Draw()
hb.Draw("same")
hb_zx = TH1F('hb_zx', 'hb_zx', 20, 0, 1)
hb_zx.SetLineColor(2)
hb_zx.Draw()
hb_zx_rw = TH1F('hb_zx', 'hb_zx', 20, 0, 1)
hb_zx_rw.SetLineColor(2)
hb_zx_rw.Draw()
hb_zx_sel = TH1F('hb_zx', 'hb_zx', 20, 0, 1)
hb_zx_sel.SetLineColor(2)
hb_zx_sel.Draw()
hmelas = TH1F('mela_s', 'mela_s', bins, 0, 1)
hmelas.SetLineColor(2)
hmelab = TH1F('mela_b', 'mela_b', bins, 0, 1)
hmelab.SetLineColor(1)
hmelas.Draw()
hmelab.Draw("same")
i = 0
c1 = TCanvas('c1', 'mva output', 200, 10, 700, 500)
c1.cd()
c1.SetGridx()
c1.SetGridy()
c1.Draw()
gStyle.SetOptStat(0)
#------- sig events ---------
stotal_ = 0
stotal_sel = 0
sigweight = 0
#do signal
for line in sig:
data = line.rsplit()
# data_ = [float(x) for x in line.rsplit()]
data_ = map(float, line.rsplit())
d = dict(zip(columns, data))
inputs = []
if (float(d["mass4l"]) < 110):
continue
if (float(d["mass4l"]) > 150):
continue
channel_ = int(d["channel"])
weight = float(d["weight"])
if (int(d["channel"]) == 1):
for var in mvavars1:
inputs.append(float(d[var]))
y = bnn_4mu_loosesel4_ext.bnn_4mu_loosesel4(inputs, 100, 200)
if (int(d["channel"]) == 2):
for var in mvavars2:
inputs.append(float(d[var]))
y = bnn_4e_loosesel4_ext.bnn_4e_loosesel4(inputs, 100, 200)
if (int(d["channel"]) == 3):
for var in mvavars3:
inputs.append(float(d[var]))
y = bnn_2e2mu_loosesel4_ext.bnn_2e2mu_loosesel4(inputs, 100, 200)
weight = float(d["weight"])
stotal_ += float(d["weight"])
if (applycuts(columns, data_, channel_)):
stotal_sel += float(d["weight"])
# continue
else:
hmelas.Fill(float(d["mela"]), weight)
hs.Fill(y, weight)
hbnnvsmela_sig.Fill(y, float(d["mela"]), weight)
if (int(d["index"]) == 6):
sigweight += weight
i += 1
if (not (i % 10000)):
print i
print y
hs.Draw()
c1.Update()
sig.close()
print "Higgs 125 weight: " + str(sigweight)
print "Integral hs: " + str(hs.Integral())
hs.Scale(1 / hs.Integral())
hmelas.Scale(1 / hmelas.Integral())
#do background
line = bkg.readline()
data = line.rsplit()
columns = []
for varname in data:
print varname
varname = varname.lstrip(':b')
columns.append(varname)
#------- ZZ events ---------
average = 0
btotal_ = 0
btotal_sel = 0
i = 0
for line in bkg:
data = line.rsplit()
# data_ = [float(x) for x in line.rsplit()]
data_ = map(float, line.rsplit())
d = dict(zip(columns, data))
channel_ = d["channel"]
if (float(d["mass4l"]) < 110):
continue
if (float(d["mass4l"]) > 150):
continue
weight = float(d["weight"])
y = 0
inputs = []
if (int(d["channel"]) == 1):
for var in mvavars1:
inputs.append(float(d[var]))
y = bnn_4mu_loosesel4_ext.bnn_4mu_loosesel4(inputs, 100, 200)
if (int(d["channel"]) == 2):
for var in mvavars2:
inputs.append(float(d[var]))
y = bnn_4e_loosesel4_ext.bnn_4e_loosesel4(inputs, 100, 200)
if (int(d["channel"]) == 3):
for var in mvavars3:
inputs.append(float(d[var]))
y = bnn_2e2mu_loosesel4_ext.bnn_2e2mu_loosesel4(inputs, 100, 200)
if (d["mela"] == "-nan"):
d["mela"] = "0"
btotal_ += float(weight)
if (applycuts(columns, data_, int(d["channel"]))):
# print "notselected"
btotal_sel += (weight)
# continue
else:
hmelab.Fill(float(d["mela"]), weight)
hb.Fill(y, weight)
hbnnvsmela_zz.Fill(y, float(d["mela"]), weight)
i += 1
if (not (i % 10000)):
print i
print y
c1.Update()
average += float(d["mass4l"])
# print "average: " + str(average/i)
print "btotal_: " + str(btotal_)
print "Integral hb: " + str(hb.Integral())
dat = open('/home/jbochenek/data/HZZ4l_2012_data/hzz_loosesel/zx_mela.dat')
line = dat.readline()
data = line.rsplit()
columns = []
for varname in data:
print varname
varname = varname.lstrip(':b')
columns.append(varname)
#------- Z+X events ---------
f_fake = TFile('../fake_reweight.root')
f_fake.cd()
cratio_Run2012_e = gDirectory.Get("cratio_Run2012_e")
cratio_Run2012_mu = gDirectory.Get("cratio_Run2012_mu")
for line in dat:
data = line.rsplit()
data_ = map(float, line.rsplit())
d = dict(zip(columns, data))
reweight = 1
inputs = []
inputs = []
if (int(d["channel"]) == 1):
for var in mvavars1:
inputs.append(float(d[var]))
y = bnn_4mu_loosesel4_ext.bnn_4mu_loosesel4(inputs, 100, 200)
reweight *= cratio_Run2012_mu.Interpolate(float(
d["lept3_eta"]), float(
d["lept3_pt"])) * cratio_Run2012_mu.Interpolate(
float(d["lept4_eta"]), float(d["lept4_pt"]))
if (int(d["channel"]) == 2):
for var in mvavars2:
inputs.append(float(d[var]))
y = bnn_4e_loosesel4_ext.bnn_4e_loosesel4(inputs, 100, 200)
reweight *= cratio_Run2012_e.Interpolate(float(
d["lept3_eta"]), float(
d["lept3_pt"])) * cratio_Run2012_e.Interpolate(
float(d["lept4_eta"]), float(d["lept4_pt"]))
if (int(d["channel"]) == 3):
for var in mvavars3:
inputs.append(float(d[var]))
y = bnn_2e2mu_loosesel4_ext.bnn_2e2mu_loosesel4(inputs, 100, 200)
reweight *= cratio_Run2012_e.Interpolate(float(
d["lept3_eta"]), float(
d["lept3_pt"])) * cratio_Run2012_e.Interpolate(
float(d["lept4_eta"]), float(d["lept4_pt"]))
weight = float(d["weight"])
d["issamesign"] = "0"
if (d["mela"] == "-nan"):
d["mela"] = "0"
btotal_ += float(d["weight"])
d["lept3_pfx"] = "0"
d["lept4_pfx"] = "0"
d["lept3_mvaid"] = "1.0"
d["lept4_mvaid"] = "1.0"
for k, v in enumerate(data_):
data_[k] = float(d[columns[k]])
# print columns[k] + " " + str(v)
if (applycuts(columns, data_, int(d["channel"]))):
# print "notselected"
btotal_sel += weight
else:
# hb_zx_sel.Fill(float(d["mela"]), 1.)
hmelab.Fill(float(d["mela"]), reweight)
hbnnvsmela_zx.Fill(y, float(d["mela"]), weight)
hb.Fill(y, reweight)
hb_zx_rw.Fill(float(d["mela"]), reweight)
hb_zx.Fill(float(d["mela"]), 1.)
i += 1
if (not (i % 10000)):
print i
print y
hb.Draw("same")
c1.Update()
average += float(d["mass4l"])
hmelas.Fill(0.0, stotal_sel)
hmelab.Fill(0.0, btotal_sel)
# hs.Fill(0.0, stotal_sel)
# hb.Fill(0.0, btotal_sel)
# print "average: " + str(average/i)
print "btotal_: " + str(btotal_)
print "Integral hb: " + str(hb.Integral())
print "Integral zx: " + str(hb_zx.Integral())
print "Integral zx (sel.): " + str(hb_zx_sel.Integral())
print "Integral zx (rw): " + str(hb_zx_rw.Integral())
#------- Print Z+X plots ---------
c_zx = TCanvas('c_zx', 'mva zx output', 200, 10, 700, 500)
c_zx.cd()
c_zx.SetGridx()
c_zx.SetGridy()
c_zx.Draw()
gStyle.SetOptStat(0)
hb_zx.Scale(1 / hb_zx.Integral())
# hb_zx_sel.Scale(1/hb_zx_sel.Integral())
hb_zx_rw.Scale(1 / hb_zx_rw.Integral())
c_zx.cd()
# c_zx.SetLogy()
hb_zx.SetTitle("BNN Discriminant: Z+X")
hb_zx.GetXaxis().SetTitle("D(x)")
legend = TLegend(0.3, 0.72, 0.68, 0.88)
hb_zx.SetLineColor(3)
hb_zx.Draw()
hb_zx_rw.SetLineColor(2)
hb_zx_rw.SetLineWidth(2)
hb_zx.SetLineWidth(2)
hb_zx_sel.SetLineWidth(2)
hb_zx_rw.Draw("same")
hb_zx_sel.SetLineColor(1)
# hb_zx_sel.Draw("same")
hb_zx.Draw("hist same")
legend.AddEntry(hb_zx_rw, "reweight", "l")
legend.AddEntry(hb_zx, "unweighted", "l")
# legend.AddEntry(hb_zx_sel, "selected", "l");
legend.Draw("same")
c_zx.Update()
c_zx.SaveAs("zx-reweight_" + evtset + ".gif")
#------- Print 2D plots ---------
c_2d = TCanvas('c_2d', 'mva zx output', 200, 10, 500, 500)
c_2d.cd()
c_2d.SetGridx()
c_2d.SetGridy()
c_2d.Draw()
gStyle.SetOptStat(0)
c_2d.cd()
hbnnvsmela_zz.SetTitle("BNN/MELA")
hbnnvsmela_zz.GetXaxis().SetTitle("BNN(x)")
hbnnvsmela_zz.GetYaxis().SetTitle("MELA(x)")
legend = TLegend(0.22, 0.75, 0.78, 0.88)
hbnnvsmela_zx.SetMarkerColor(3)
# hbnnvsmela_zx.Draw()
hbnnvsmela_zz.SetMarkerColor(2)
hbnnvsmela_sig.SetMarkerColor(4)
hbnnvsmela_zz.SetFillColor(2)
hbnnvsmela_sig.SetFillColor(4)
hbnnvsmela_zx.SetFillColor(3)
hbnnvsmela_zz.SetLineWidth(2)
hbnnvsmela_zx.SetLineWidth(2)
hbnnvsmela_sig.SetLineWidth(2)
hbnnvsmela_zz.SetMarkerSize(2.)
hbnnvsmela_sig.SetMarkerSize(2.)
hbnnvsmela_zz.Draw("box")
# hbnnvsmela_sig.SetMarkerColor(1)
hbnnvsmela_sig.Draw("same box")
legend.AddEntry(
hbnnvsmela_zz,
"ZZ: corr. " + str(round(hbnnvsmela_zz.GetCorrelationFactor(), 3)),
"p")
# legend.AddEntry(hbnnvsmela_zx, "Z+X", "p");
legend.AddEntry(
hbnnvsmela_sig,
"sig.: corr. " + str(round(hbnnvsmela_sig.GetCorrelationFactor(), 3)),
"p")
legend.Draw("same")
c_2d.Update()
c_2d.SaveAs("bnn_vs_mela_" + evtset + ".gif")
line = dat.readline()
data = line.rsplit()
columns = []
for varname in data:
print varname
varname = varname.lstrip(':b')
columns.append(varname)
print "btotal_: " + str(btotal_)
print "Integral hb: " + str(hb.Integral())
bkg.close()
#------- BNN eff ---------
geff = TGraph()
geff.SetName("bnneff")
gmelaeff = TGraph()
gmelaeff.SetName("melaeff")
stotal = hs.Integral()
btotal = hb.Integral()
cs = cdf(hs)
cb = cdf(hb)
maxsig = 0
optmvax = 0
optmvay = 0
for i in range(hs.GetNbinsX()):
cs1 = stotal - cs[i - 1]
cb1 = btotal - cb[i - 1]
es = cs1 / stotal
eb = cb1 / btotal
sign = 0
if (cb1 > 0):
sign = cs1 / sqrt(cs1 + cb1)
if (sign > maxsig):
maxsig = sign
optcut = i
optmvax = eb
optmvay = es
heff.Fill(eb, es)
if (i < 2):
continue
else:
geff.SetPoint(geff.GetN(), eb, es)
print "\n Maximum Significance " + str(maxsig) + ", is bin: " + str(
optcut) + ", which is BNN Cut: " + str(optcut / bins)
print "es: " + str(optmvay) + "\teb: " + str(optmvax)
#------- MELA eff ---------
smelatotal = hmelas.Integral()
bmelatotal = hmelab.Integral()
cmelas = cdf(hmelas)
cmelab = cdf(hmelab)
maxsig = 0
optmvax = 0
optmvay = 0
for i in range(hmelas.GetNbinsX()):
cs1 = smelatotal - (cmelas[i - 1])
cb1 = bmelatotal - (cmelab[i - 1])
es = cs1 / smelatotal
eb = cb1 / bmelatotal
sign = 0
if (cb1 > 0):
sign = cs1 / sqrt(cs1 + cb1)
if (sign > maxsig):
maxsig = sign
optcut = i
optmvax = eb
optmvay = es
heffmela.Fill(eb, es)
if (i < 2):
continue
else:
gmelaeff.SetPoint(gmelaeff.GetN(), eb, es)
# print str(i) + " eb: " + str(eb) + " es: " + str(es)
print "\n Maximum Significance " + str(maxsig) + ", is bin: " + str(
optcut) + ", which is MELA Cut: " + str(optcut / bins)
print "es: " + str(optmvay) + "\teb: " + str(optmvax)
print "stotal: " + str(stotal_)
print "btotal: " + str(btotal_)
print "stotal_sel: " + str(stotal_sel)
print "btotal_sel: " + str(btotal_sel)
ceffmela = TCanvas('ceffmela', 'effmela', 200, 10, 700, 500)
ceffmela.cd()
gStyle.SetOptStat(0)
heffmela.SetMarkerSize(2)
ceffmela.Update()
ceffmela.SaveAs("heffmela_" + evtset + ".gif")
ceffmela.Update()
ceff = TCanvas('ceff', 'eff', 200, 10, 700, 500)
ceff.cd()
legend_x = TLegend(0.6, 0.25, 0.85, 0.45)
gStyle.SetOptStat(0)
heff.SetLineColor(2)
geff.SetLineColor(2)
geff.SetLineWidth(2)
heff.SetTitle("ROC for BNN & MELA (bkg = ZZ & Z+X)")
heff.GetXaxis().SetTitle("#epsilon_{bkg.}")
heff.GetXaxis().SetTitleSize(0.06)
heff.GetYaxis().SetTitleSize(0.06)
heff.GetXaxis().SetTitleOffset(0.65)
heff.GetYaxis().SetTitleOffset(0.65)
heff.GetYaxis().SetTitle("#epsilon_{sig.}")
heff.Draw()
geff.Draw("same")
heffmela.SetLineColor(3)
gmelaeff.SetLineColor(3)
gmelaeff.SetLineWidth(2)
heffmela.Draw("same")
gmelaeff.Draw("same")
legend_x.AddEntry(heffmela, "MELA", "l")
legend_x.AddEntry(heff, "BNN", "l")
legend_x.Draw("same")
ceff.Update()
ceff.SaveAs("heff_" + evtset + ".gif")
ceff.Update()
#------- BNN Dist ---------
# c1.SetLogy();
hb.Scale(1 / hb.Integral())
hmelab.Scale(1 / hmelab.Integral())
c1.cd()
gStyle.SetOptStat(0)
hs.Rebin(2)
hb.Rebin(2)
hs.SetTitle("BNN Discriminant: sig. & bkg (ZZ & Z+X)")
hs.GetXaxis().SetTitle("D(x)")
hs.SetLineWidth(2)
hb.SetLineWidth(2)
hs.GetXaxis().SetTitleSize(0.06)
hs.GetYaxis().SetTitleSize(0.06)
hs.GetXaxis().SetTitleOffset(0.65)
hs.GetYaxis().SetTitleOffset(0.65)
hs.Draw()
hb.Draw("same")
c1.Update()
c1.SaveAs("hdist_bnn_" + evtset + ".gif")
#------- MELA Dist ---------
c2 = TCanvas('c2', 'mva output', 200, 10, 700, 500)
c2.cd()
gStyle.SetOptStat(0)
c2.SetGridx()
c2.SetGridy()
c2.Draw()
# c2.SetLogy();
# hmelas.Scale(1/hmelas.Integral())
# hmelab.Scale(1/hmelab.Integral())
gStyle.SetOptStat(0)
hmelas.Rebin(2)
hmelab.Rebin(2)
hmelab.SetTitle("MELA Discriminant: sig. & bkg (ZZ & Z+X)")
hmelab.GetXaxis().SetTitle("D(x)")
hmelab.SetLineWidth(2)
hmelas.SetLineWidth(2)
hmelab.GetXaxis().SetTitleSize(0.06)
hmelab.GetYaxis().SetTitleSize(0.06)
hmelab.GetXaxis().SetTitleOffset(0.65)
hmelab.GetYaxis().SetTitleOffset(0.65)
hmelab.Draw()
hmelas.Draw("same")
c2.Update()
c2.SaveAs("hdist_mela_" + evtset + ".gif")