if obj.InheritsFrom("TH1") and signalMatch in name:
if name.split('_')[-2] == "group":
if "outliers" in name:
drawThisBin = True
name2D = 'W{ch}_{flav}_outliers'.format(
ch=charge, flav=channel)
title2D = 'W{chs}: |#eta| > {etamax}#; p_{{T}} #notin [{ptmin:.0f},{ptmax:.0f})'.format(
etamax=genBins.etaBins[-1],
ptmin=genBins.ptBins[0],
ptmax=genBins.ptBins[-1],
chs=chs)
else:
etabinIndex, ptbinIndex = get_ieta_ipt_from_process_name(
name)
if options.draw_all_bins: drawThisBin = True
elif options.draw_selected_etaPt != '':
if etabinIndex != ieta_sel or ptbinIndex != ipt_sel:
drawThisBin = False
else:
drawThisBin = True
else:
drawThisBin = False
name2D = 'W{ch}_{flav}_ieta_{ieta}_ipt_{ipt}'.format(
ch=charge,
flav=channel,
ieta=etabinIndex,
ipt=ptbinIndex)
title2D = 'W{chs}: |#eta| #in [{etamin},{etamax})#; p_{{T}} #in [{ptmin:.0f},{ptmax:.0f})'.format(
etamin=genBins.etaBins[etabinIndex],
pois_indices[impMat.GetXaxis().GetBinLabel(ib)] = ib
nuisances = []
nuisances_indices = []
for ib in xrange(1, impMat.GetNbinsY() + 1):
for n in nuis_regexps:
if re.match(n, impMat.GetYaxis().GetBinLabel(ib)):
nuisances.append(impMat.GetYaxis().GetBinLabel(ib))
nuisances_indices.append(ib)
mat = {}
pois = sorted(pois,
key=lambda x: int(x.split('_')[-2]) if '_Ybin_' in x else 0)
pois = sorted(pois,
key=lambda x: get_ieta_ipt_from_process_name(x)
if ('_ieta_' in x and '_ipt_' in x) else 0)
# sort by charge if needed
pois = sorted(pois,
key=lambda x: 0 if "plus" in x else 1 if "minus" in x else 2)
for ipoi, poi in enumerate(pois):
for inuis, nuis in enumerate(nuisances):
mat[(poi, nuis)] = impMat.GetBinContent(pois_indices[poi],
nuisances_indices[inuis])
## sort the pois and nuisances alphabetically, except for pdfs, which are sorted by number
if len(options.nuisgroups) == 0:
## for mu* QCD scales, distinguish among muR and muRXX with XX in 1-10
nuisances = sorted(nuisances,
key=lambda x: int(x.replace('pdf', ''))
th2_sub.GetZaxis().SetTitle("impact for {p} {units}".format(
units='' if options.absolute else '(%)',
p=poiName_target[options.target]))
th2_sub.GetZaxis().SetTitleOffset(1.4)
th2_sub.GetZaxis().SetTitleSize(0.05)
th2_sub.GetZaxis().SetLabelSize(0.04)
for p in pois:
if charge != "allCharges" and not charge in p: continue
if '_ieta_' in p and '_ipt_' in p:
if options.absolute:
val = mat[(p, nuisance[n])]
else:
val = 100 * mat[(p, nuisance[n])] / valuesAndErrors[p][
0] if valuesAndErrors[p][0] != 0 else 0.0
ieta, ipt = get_ieta_ipt_from_process_name(p)
th2_sub.SetBinContent(ieta + 1, ipt + 1, val)
rmax = max(abs(th2_sub.GetMaximum()), abs(th2_sub.GetMinimum()))
if not options.absolute: rmax = min(20., rmax)
if options.absValue:
th2_sub.GetZaxis().SetRangeUser(0, rmax)
else:
th2_sub.GetZaxis().SetRangeUser(-rmax, rmax)
if options.zrange:
if options.zrange == "template":
rmin = th2_sub.GetBinContent(th2_sub.GetMinimumBin())
rmax = th2_sub.GetBinContent(th2_sub.GetMaximumBin())
else:
rmin, rmax = (float(x) for x in options.zrange.split(','))
th2_sub.GetZaxis().SetRangeUser(rmin, rmax)
def niceName(name, genEtaPtBins="", forceLep="", drawRangeInLabel=False):
if '_Ybin' in name:
nn = '#mu: ' if '_mu_' in name else 'el: ' if '_el_' in name else 'l: '
if 'plus' in name: nn += 'W+ '
elif 'minus' in name: nn += 'W- '
else: nn += 'W '
if 'left' in name: nn += 'left '
elif 'right' in name: nn += 'right '
elif 'long' in name: nn += 'long '
else: nn += 'unpolarized '
idx = -2 if (name.endswith('mu') or any(
[x in name
for x in ['masked', 'sumxsec', 'charge', 'a0', 'a4']])) else -1
nn += name.split('_')[idx]
if 'eff_unc' in name:
nn = '#epsilon_{unc}^{' + nn + '}'
return nn
elif '_ieta_' in name and '_ipt_' in name:
nn = '#mu: ' if '_mu_' in name else 'el: ' if '_el_' in name else 'l:'
nn += 'W+ ' if 'plus' in name else 'W- ' if 'minus' in name else 'W '
ieta, ipt = get_ieta_ipt_from_process_name(name)
nn += "i#eta, ip_{{T}} = {neta}, {npt} ".format(neta=ieta, npt=ipt)
return nn
elif '_ieta_' in name:
# nn = '#mu: ' if '_mu_' in name else 'el: ' if '_el_' in name else 'l:'
# nn += 'W+ ' if 'plus' in name else 'W- ' if 'minus' in name else 'W '
ieta = int((name.split("_ieta_")[1]).split("_")[0])
# nn += "i#eta = {neta}".format(neta=ieta)
nn = ""
if drawRangeInLabel:
etal, etah = 0.0, 0.0
if genEtaPtBins:
etal = genEtaPtBins.etaBins[ieta]
etah = genEtaPtBins.etaBins[ieta + 1]
wch = 'W+ ' if 'plus' in name else 'W- ' if 'minus' in name else 'W '
lep = '#mu' if '_mu_' in name else 'el' if '_el_' in name else 'l'
if forceLep:
lep = '#mu' if forceLep == "mu" else 'e' if forceLep == "el" else 'l'
nn = "{wch} |#eta^{{{lep}}}| #in [{etal:1.1f}, {etah:1.1f}]".format(
wch=wch, lep=lep, etal=etal, etah=etah)
else:
nn = '#mu: ' if '_mu_' in name else 'el: ' if '_el_' in name else 'l:'
nn += 'W+ ' if 'plus' in name else 'W- ' if 'minus' in name else 'W '
nn += "i#eta = {neta}".format(neta=ieta)
return nn
elif '_ipt_' in name:
# nn = '#mu: ' if '_mu_' in name else 'el: ' if '_el_' in name else 'l:'
# nn += 'W+ ' if 'plus' in name else 'W- ' if 'minus' in name else 'W '
ipt = int((name.split("_ipt_")[1]).split("_")[0])
# nn += "ip_{{T}} = {npt}".format(npt=ipt)
nn = ""
if drawRangeInLabel:
ptl, pth = 0.0, 0.0
if genEtaPtBins:
ptl = genEtaPtBins.ptBins[ipt]
pth = genEtaPtBins.ptBins[ipt + 1]
wch = 'W+ ' if 'plus' in name else 'W- ' if 'minus' in name else 'W '
lep = '#mu' if '_mu_' in name else 'el' if '_el_' in name else 'l'
if forceLep:
lep = '#mu' if forceLep == "mu" else 'e' if forceLep == "el" else 'l'
nn = "{wch} p_{{T}}({lep}) #in [{ptl:3g}, {pth:3g}]".format(
wch=wch, lep=lep, ptl=ptl, pth=pth)
else:
nn = '#mu: ' if '_mu_' in name else 'el: ' if '_el_' in name else 'l:'
nn += 'W+ ' if 'plus' in name else 'W- ' if 'minus' in name else 'W '
nn += "ip_{{T}} = {npt}".format(npt=ipt)
return nn
elif "CMS_" in name:
# keep Wmu or We now that we do combination, they are different sources
# if "CMS_Wmu" in name:
# return name.replace("CMS_Wmu_","")
# elif "CMS_We" in name:
# return name.replace("CMS_We_","")
#else:
# return name
return name.replace("CMS_", "")
elif re.match("Fakes(Eta|EtaCharge|PtNorm|PtSlope)Uncorrelated.*", name):
num = re.findall(r'\d+', name) # get number
pfx = name.split(
num[0])[1] # split on number and read what's on the right
leptonCharge = ""
if len(pfx):
leptonCharge = "{lep}{chs}".format(
lep="#mu" if "mu" in pfx else "e",
chs="+" if "plus" in pfx else "-" if "minus" in pfx else "")
tmpvar = ""
if "FakesEtaCharge" in name: tmpvar = "#eta-byCharge"
elif "FakesEta" in name: tmpvar = "#eta"
elif "FakesPtNorm" in name: tmpvar = "p_{T}-norm"
elif "FakesPtSlope" in name: tmpvar = "p_{T}-slope"
return "Fakes {var}-{n} {lepCh}".format(var=tmpvar,
n=num[0],
lepCh=leptonCharge)
elif re.match(".*EffStat\d+.*", name):
num = re.findall(
r'\d+',
name) # get number (there will be two of them, need the second)
pfx = name.split("EffStat" + str(num[1]))[
1] # split on second number and read what's on the right
leptonCharge = ""
if len(pfx):
leptonCharge = "{lep}{chs}".format(
lep="#mu" if "mu" in pfx else "e",
chs="+" if "plus" in pfx else "-" if "minus" in pfx else "")
return "Eff. uncorr. {n1}-{n2} {lepCh}".format(n1=num[0],
n2=num[1],
lepCh=leptonCharge)
else:
return name
reverse=False)
elif any(re.match('FakesPtUncorrelated.*', x) for x in params):
params = sorted(params,
key=lambda x: utilities.getNFromString(x),
reverse=False)
elif any(re.match('.*EffStat.*', x) for x in params):
params = sorted(params,
key=lambda x: utilities.getNFromString(x),
reverse=False)
elif any('masked' in x or x.endswith('mu') for x in params):
if any('_ieta_' in x for x in params):
# differential xsection: get ieta, ipt index and use them as keys to sort
params = sorted(
params,
key=lambda x: -1
if "_ieta_" not in x else get_ieta_ipt_from_process_name(x),
reverse=False)
else:
params = sorted(params,
key=lambda x: int(x.split('_')[-2])
if ('masked' in x or x.endswith('mu')) else -1,
reverse=False)
nuis_p_i = 0
title = "#theta"
#hist_fit_s = ROOT.TH1F("fit_s" ,"S+B fit Nuisances ;;%s "%title,len(params),0,len(params))
hist_fit_s = ROOT.TH1F("fit_s", '', len(params), 0, len(params))
pmin, pmax = -3., 3.
hist_fit_1d = ROOT.TH1F("fit_1d ", '', 50, pmin, pmax)
hist_fit_1d_e = ROOT.TH1F("fit_1d_e", '', 50, pmin, pmax)
def getXsecs_etaPt(processes, systs, etaPtBins, infile): # in my case here, the histograms already have the cross section in pb, no need to divide by lumi
#print "Inside getXsecs_etaPt"
# etaPtBins is a list of 4 things: Netabins, etabins, Nptbins,ptbins
histo_file = ROOT.TFile(infile, 'READ')
hists = []
for process in processes:
etabins = etaPtBins[1]
ptbins = etaPtBins[3]
charge = "plus" if "plus" in process else "minus"
# check if gen eta binning starts from negative value (probably we will stick to |eta|, but just in case)
etavar = "absetal1" if (float(etabins[0]) >= 0.0) else "etal1"
cen_name = 'gen_ptl1_'+etavar+'_W'+charge+'_mu_central'
cen_hist = histo_file.Get(cen_name) # this is a TH2
useEleXsec = False
if not cen_hist:
cen_name = cen_name.replace("_mu_","_el_")
cen_hist = histo_file.Get(cen_name)
useEleXsec = True
#print cen_name
if not cen_hist:
print "Error in getXsecs_etaPt(): histogram %s not found in file %s. Exit" % (cen_name, infile)
quit()
# before searching the bin, sum epsilon to the y value being inspected
# Root assigns the lower bin edge to the bin, while the upper bin edge is assigned to the adjacent bin. However, depending on the number y being used,
# there can be a precision issue which might induce the selection of the wrong bin (since yfirst and yvalue are actually bin boundaries)
# It seems odd, but I noticed that with the fake rate graphs (I was getting events migrating between adjacent eta bins)
epsilon = 0.00001
# process has the form: Wplus_el_ieta_3_ipt_0_Wplus_el_group_0, where the number after group is not relevant (does not coincide with absolute bin number)
# as it can be the same for many processes
# one should use ieta and ipt, which identifies the template bin (first index is 0, not 1)
# generally, ieta,ipt in name should correspond to 2D histogram X and Y bins (after adding 1)
# However, one could use different definition for the gen level binning (used to cut and define a signal process) and the reco level one (defining the TH2)
if "outliers" in process:
# cross section for outliers = Integral(all) - integral(acceptance)
totxsec = cen_hist.Integral(0, cen_hist.GetNbinsX()+1, 0, cen_hist.GetNbinsY()+1)
ieta_low_acc = cen_hist.GetXaxis().FindFixBin(etabins[0]+epsilon)
ieta_high_acc = cen_hist.GetXaxis().FindFixBin(etabins[-1]+epsilon) -1
ipt_low_acc = cen_hist.GetYaxis().FindFixBin(ptbins[0]+epsilon)
ipt_high_acc = cen_hist.GetYaxis().FindFixBin(ptbins[-1]+epsilon) -1
xsecAccept = cen_hist.Integral(ieta_low_acc, ieta_high_acc, ipt_low_acc, ipt_high_acc)
ncen = totxsec - xsecAccept
else:
ieta,ipt = get_ieta_ipt_from_process_name(process)
etafirst = etabins[ieta]
etalast = etabins[ieta+1]
ptfirst = ptbins[ipt]
ptlast = ptbins[ipt+1]
# caution, we are using TH2, the logic below is slightly different than for TH1
istart_eta = cen_hist.GetXaxis().FindFixBin(etafirst + epsilon)
iend_eta = cen_hist.GetXaxis().FindFixBin(etalast + epsilon)
istart_pt = cen_hist.GetYaxis().FindFixBin(ptfirst + epsilon)
iend_pt = cen_hist.GetYaxis().FindFixBin(ptlast + epsilon)
ncen = cen_hist.Integral(istart_eta, iend_eta-1, istart_pt,iend_pt-1)
tmp_hist = ROOT.TH1F('x_'+process,'x_'+process, 1, 0., 1.)
## normalize back to cross section
tmp_hist.SetBinContent(1, ncen)
hists.append(copy.deepcopy(tmp_hist))
for sys in systs:
upn = sys+'Up' if not 'pdf' in sys else sys
dnn = sys+'Dn' if not 'pdf' in sys else sys
if useEleXsec:
sys_upname = 'gen_ptl1_'+etavar+'_W'+charge+'_el_'+upn
sys_dnname = 'gen_ptl1_'+etavar+'_W'+charge+'_el_'+dnn
else:
sys_upname = 'gen_ptl1_'+etavar+'_W'+charge+'_mu_'+upn
sys_dnname = 'gen_ptl1_'+etavar+'_W'+charge+'_mu_'+dnn
sys_up_hist = histo_file.Get(sys_upname)
sys_dn_hist = histo_file.Get(sys_dnname)
if not sys_up_hist:
print "Error in getXsecs_etaPt(): histogram %s not found in file %s" % (sys_upname, infile)
quit()
if not sys_dn_hist:
print "Error in getXsecs_etaPt(): histogram %s not found in file %s" % (sys_dnname, infile)
quit()
if "outliers" in process:
totxsec = sys_up_hist.Integral(0, sys_up_hist.GetNbinsX()+1, 0, sys_up_hist.GetNbinsY()+1)
xsecAccept = sys_up_hist.Integral(ieta_low_acc, ieta_high_acc, ipt_low_acc, ipt_high_acc)
nup = totxsec - xsecAccept
totxsec = sys_dn_hist.Integral(0, sys_dn_hist.GetNbinsX()+1, 0, sys_dn_hist.GetNbinsY()+1)
xsecAccept = sys_dn_hist.Integral(ieta_low_acc, ieta_high_acc, ipt_low_acc, ipt_high_acc)
ndn = totxsec - xsecAccept
else:
nup = sys_up_hist.Integral(istart_eta, iend_eta-1, istart_pt,iend_pt-1)
ndn = sys_dn_hist.Integral(istart_eta, iend_eta-1, istart_pt,iend_pt-1)
if 'pdf' in sys:
ndn = 2.*ncen-nup ## or ncen/nup?
tmp_hist_up = ROOT.TH1F('x_'+process+'_'+sys+'Up','x_'+process+'_'+sys+'Up', 1, 0., 1.)
tmp_hist_up.SetBinContent(1, nup)
tmp_hist_dn = ROOT.TH1F('x_'+process+'_'+sys+'Down','x_'+process+'_'+sys+'Dn', 1, 0., 1.)
tmp_hist_dn.SetBinContent(1, ndn)
hists.append(copy.deepcopy(tmp_hist_up))
hists.append(copy.deepcopy(tmp_hist_dn))
hist_data = ROOT.TH1F('x_data_obs', 'x_data_obs', 1, 0., 1.)
hist_data.SetBinContent(1, 1.)
hists.append(copy.deepcopy(hist_data))
return hists
def plotPars(inputFile, pois=None, selectString='', maxPullsPerPlot=30, plotdir='./', suffix='', analysis='helicity',
excludeName=None, paramFamily=None,plotPull=False,
channel='el', charge='plus',
selection="",useMedian=False):
chs = "+" if charge == "plus" else "-"
isSignalStrength = False
if paramFamily:
if paramFamily == "pdf":
pois="pdf.*"
all_valuesAndErrors = utilities.getHistosFromToys(inputFile,getPull=plotPull,matchBranch=pois,excludeBranch=excludeName,
selection=selection,setStatOverflow=True,getMedian=useMedian)
elif paramFamily == "scale":
pois=""
for i in range(1,11):
pois += "muR{n},muF{n},muRmuF{n},".format(n=str(i))
if pois.endswith(','): pois = pois[:-1]
all_valuesAndErrors = utilities.getHistosFromToys(inputFile,getPull=plotPull,matchBranch=pois,excludeBranch=excludeName,
selection=selection,setStatOverflow=True,getMedian=useMedian)
elif paramFamily == "shapesyst":
if channel == "el":
pois = "alphaS,CMS_We_sig_lepeff,CMS_We_elescale,CMS_We_FRe_slope,CMS_We_FRe_continuous"
else:
pois = "alphaS,CMS_Wmu_sig_lepeff,CMS_Wmu_muscale,CMS_Wmu_FRmu_slope,CMS_Wmu_FR_continuous"
all_valuesAndErrors = utilities.getHistosFromToys(inputFile,getPull=plotPull,matchBranch=pois,
excludeBranch=excludeName,selection=selection,setStatOverflow=True,getMedian=useMedian)
elif paramFamily == "normsyst":
if channel == "el":
pois = "CMS_We_lepVeto,CMS_We_bkg_lepeff,CMS_VV,CMS_Top,CMS_W,CMS_DY,CMS_lumi_13TeV,CMS_We_flips,CMS_Wbkg"
else:
pois = "CMS_Wmu_lepVeto,CMS_Wmu_bkg_lepeff,CMS_VV,CMS_Top,CMS_W,CMS_DY,CMS_lumi_13TeV,CMS_Wbkg"
all_valuesAndErrors = utilities.getHistosFromToys(inputFile,getPull=plotPull,matchBranch=pois,
excludeBranch=excludeName,selection=selection,setStatOverflow=True,getMedian=useMedian)
elif paramFamily == "signalStrength":
# add some exclusion matches if not already present
if excludeName != None: excludeName += ",.*pmaskedexp.*" if "pmaskedexp" not in excludeName else ""
else : excludeName = ".*pmaskedexp.*"
pois="W{ch}.*_mu".format(ch=charge)
isSignalStrength = True
all_valuesAndErrors = utilities.getHistosFromToys(inputFile,200,0,2,getPull=plotPull,matchBranch=pois,
excludeBranch=excludeName,selection=selection,setStatOverflow=True,getMedian=useMedian)
elif paramFamily == "absxsec":
pois="W{ch}.*_pmaskedexp".format(ch=charge)
if excludeName != None: excludeName += ",.*pmaskedexpnorm.*" if "pmaskedexpnorm" not in excludeName else ""
else : excludeName = ".*pmaskedexpnorm.*"
all_valuesAndErrors = utilities.getHistosFromToys(inputFile,1000,0,200,getPull=plotPull,matchBranch=pois,
excludeBranch=excludeName,selection=selection,setStatOverflow=True,getMedian=useMedian)
elif paramFamily == "normxsec":
pois="W{ch}.*_pmaskedexpnorm".format(ch=charge)
all_valuesAndErrors = utilities.getHistosFromToys(inputFile,1000,0,1,getPull=plotPull,matchBranch=pois,
excludeBranch=excludeName,selection=selection,setStatOverflow=True,getMedian=useMedian)
else:
# warning: utilities.getHistosFromToys is working only for parameters centered around 0 (histogram is defined in -3,3)
# this script will not work if using parameters with pmaskedexp
if pois:
if any (["pmaskedexp" in x for x in pois.split(",")]):
print "Warning: current setup cannot work with parameters matching pmaskedexp: they are expected to be centered around 0. Exit"
print "You might want to change behaviour of function utilities.getHistosFromToys to use those parameters as well"
exit(0)
if any ([wc in poi for wc in ["Wplus","Wminus"] for poi in pois.split(",")]):
all_valuesAndErrors = utilities.getHistosFromToys(inputFile,200,0,2,getPull=plotPull,matchBranch=pois,
excludeBranch=excludeName,selection=selection,setStatOverflow=True,getMedian=useMedian)
if not any("pmaskedexp" in poi for poi in pois.split(",")):
isSignalStrength = True
else:
all_valuesAndErrors = utilities.getHistosFromToys(inputFile,getPull=plotPull,matchBranch=pois,
excludeBranch=excludeName,selection=selection,setStatOverflow=True,getMedian=useMedian)
print "From the list of parameters it seems that you are plotting results for channel ",channel
print "pois = ", pois
valuesAndErrors = {}
excludedNames = excludeName.split(',') if excludeName else []
if pois:
poi_patts = pois.split(",")
for ppatt in poi_patts:
for (k,v) in all_valuesAndErrors.iteritems():
if any([re.match(x,k) for x in excludedNames]): continue
if re.match(ppatt,k) and k not in valuesAndErrors: valuesAndErrors[k] = v
#print valuesAndErrors.keys()
params = valuesAndErrors.keys()
print "----------------------------------"
print "List of parameters:"
print params
print "----------------------------------"
if any(re.match('pdf.*',x) for x in params):
params = sorted(params, key = lambda x: int(x.split('pdf')[-1]), reverse=False)
nPulls = 0
pullLabelSize = 0.028
pullSummaryMap = {}
c = ROOT.TCanvas("c","",960,800)
c.SetTickx(1)
c.SetTicky(1)
c.cd()
c.SetLeftMargin(0.14)
c.SetRightMargin(0.06)
c.SetTopMargin(0.05)
c.SetBottomMargin(0.2)
c.cd()
if isSignalStrength:
chi2title = "#chi^{2} for signal strength parameters"
chi2distr = ROOT.TH1D("chi2distr",chi2title,20,0.0,2.0)
for name in params:
print "Making pull for parameter ",name
# look in utilities.py for getHistosFromToys(: if plotting pull, the histogram is defined as (x-x_gen)x_err,
# while valuesAndErrors[name][0] and valuesAndErrors[name][1] are defined as mean and rms of original histogram from x
# if option getPull of getHistosFromToys is False, then there is no difference
mean_p, sigma_p = (valuesAndErrors[name][0],valuesAndErrors[name][1]-valuesAndErrors[name][0])
#mean_p, sigma_p = (valuesAndErrors[name][3].GetMean(),valuesAndErrors[name][3].GetStdDev())
histo = valuesAndErrors[name][3]
if isSignalStrength and histo.GetStdDev() > 0.05:
histo.Rebin(2)
histo.Draw()
histo.GetXaxis().SetTitle(histo.GetTitle());
histo.GetYaxis().SetTitle("number of toys (%d total)" % histo.Integral());
histo.SetTitle("")
histo.GetYaxis().SetTitleOffset(1.05);
histo.GetXaxis().SetTitleOffset(0.9);
histo.GetYaxis().SetTitleSize(0.05);
histo.GetXaxis().SetTitleSize(0.05);
histo.GetXaxis().SetTitle(name)
fitPull = histo.Integral()>0
if fitPull:
histo.Fit("gaus","Q")
fit = histo.GetFunction("gaus")
fit.SetLineColor(4)
lat.DrawLatex(0.16, 0.8, 'mean: {me:.3f}'.format(me=fit.GetParameter(1)))
lat.DrawLatex(0.16, 0.7, 'err : {er:.2f}'.format(er=fit.GetParameter(2)))
lat.DrawLatex(0.16, 0.6, 'chi2/ndf: {cn:.2f}'.format(cn=fit.GetChisquare()/fit.GetNDF() if fit.GetNDF()>0 else 999))
lat.DrawLatex(0.65, 0.8, 'mean: {me:.3f}'.format(me=histo.GetMean()))
lat.DrawLatex(0.65, 0.7, 'RMS : {er:.2f}'.format(er=histo.GetStdDev()))
if useMedian:
lat.DrawLatex(0.65, 0.6, 'median: {me:.3f}'.format(me=mean_p)) # in this case mean_p stores the median
if isSignalStrength:
if fit.GetParameter(2) < 0.1:
histo.GetXaxis().SetRangeUser(0.5,1.5)
if fit.GetParameter(2) < 0.05:
histo.GetXaxis().SetRangeUser(0.8,1.2)
if fit.GetNDF()>0:
chi2distr.Fill(fit.GetChisquare()/fit.GetNDF())
os.system('cp /afs/cern.ch/user/g/gpetrucc/php/index.php '+plotdir)
distrdir = plotdir+'/pulldistr'
createPlotDirAndCopyPhp(distrdir)
for ext in ['png', 'pdf']:
c.SaveAs("{pdir}/{name}_postfit_{ch}_{channel}{suffix}.{ext}".format(pdir=distrdir,name=name,ch=charge,
suffix=("_"+suffix) if len(suffix) else "",
channel=channel,ext=ext))
if fitPull:
# tlatex = ROOT.TLatex(); tlatex.SetNDC();
# tlatex.SetTextSize(0.11)
# tlatex.SetTextColor(4);
# tlatex.DrawLatex(0.65,0.80,"Mean : %.3f #pm %.3f" % (histo.GetFunction("gaus").GetParameter(1),histo.GetFunction("gaus").GetParError(1)))
# tlatex.DrawLatex(0.65,0.66,"Sigma : %.3f #pm %.3f" % (histo.GetFunction("gaus").GetParameter(2),histo.GetFunction("gaus").GetParError(2)))
# tlatex.SetTextSize(0.11);
# tlatex.SetTextColor(1);
# tlatex.DrawLatex(0.65,0.33,"Post-fit #pm #sigma_{#theta}: %.3f #pm %.3f" % (mean_P p), sigma_p))
pullSummaryMap[name]=(histo.GetFunction("gaus").GetParameter(1),histo.GetFunction("gaus").GetParameter(2),
histo.GetMean(),utilities.effSigma(histo))
nPulls += 1
if nPulls>0:
if isSignalStrength:
canvas_Chi2 = ROOT.TCanvas("canvas_Chi2","",700,700)
canvas_Chi2.SetTickx(1)
canvas_Chi2.SetTicky(1)
chi2distr.Draw("Hist")
chi2distr.GetXaxis().SetTitle("#chi^{2}")
chi2distr.GetXaxis().SetTitleOffset(0.9)
chi2distr.GetXaxis().SetTitleSize(0.05)
chi2distr.GetYaxis().SetTitle("Events")
chi2distr.GetYaxis().SetTitleOffset(1.05)
chi2distr.GetYaxis().SetTitleSize(0.05)
if chi2distr.GetEntries():
overflow = 100. * chi2distr.GetBinContent(chi2distr.GetNbinsX() + 1) / chi2distr.GetEntries() # use as XX %, multiplying by 100
else:
overflow = 0
lat.DrawLatex(0.12, 0.8, 'W{ch} #rightarrow {lep}#nu'.format(ch=chs,lep="e" if channel == "el" else "#mu"))
lat.DrawLatex(0.12, 0.7, 'mean: {me:.3f}'.format(me=chi2distr.GetMean()))
lat.DrawLatex(0.12, 0.6, 'RMS : {er:.2f}'.format(er=chi2distr.GetStdDev()))
lat.DrawLatex(0.12, 0.5, 'Overflow : {of:.1f}%'.format(of=overflow))
# following does not work, the stat box is disabled
# chi2distr.SetStats(1)
# canvas_Chi2.Update()
# statBox = chi2distr.FindObject("stats")
# if statBox:
# statBox.SetX1NDC(0.62)
# statBox.SetX2NDC(0.92)
# statBox.SetY1NDC(0.59)
# statBox.SetY2NDC(0.91)
# statBox.SetFillColor(0)
# statBox.SetFillStyle(0)
# statBox.SetBorderSize(0)
# statBox.Draw()
# else:
# print "Warning: no stat box found for Chi^2"
for ext in ['png', 'pdf']:
canvas_Chi2.SaveAs("{pdir}/chi2_{ch}_{channel}{suffix}.{ext}".format(pdir=distrdir,ch=charge,
suffix=("_"+suffix) if len(suffix) else "",
channel=channel,ext=ext))
print "Generating Pull Summaries...\n"
nRemainingPulls = nPulls
hc = ROOT.TCanvas("hc","",3000,2000);
hc.SetTopMargin(0.05)
hc.SetBottomMargin(0.2)
hc.SetRightMargin(0.05)
hc.SetGrid(0);
pullPlots = 1;
while nRemainingPulls > 0:
nThisPulls = min(maxPullsPerPlot,nRemainingPulls)
pull_rms = ROOT.TH1F("pull_rms_{pp}".format(pp=str(pullPlots)) ,"", 3*nThisPulls+1,0,nThisPulls*3+1);
pull_effsigma = ROOT.TH1F("pull_effsigma_{pp}".format(pp=str(pullPlots)) ,"", 3*nThisPulls+1,0,nThisPulls*3+1);
pull_median = ROOT.TH1F("pull_median_{pp}".format(pp=str(pullPlots)) ,"", 3*nThisPulls+1,0,nThisPulls*3+1);
pi=1
sortedpulls = []
if 'pdf' in pois:
sortedpulls = sorted(pullSummaryMap.keys(), key=lambda k: int(k.split('pdf')[-1]))
# following condition should not happen (parameter is not centered around 0)
#elif 'masked' in pois:
#elif any([x in pois for x in ["Wplus","Wminus"]]) and not "masked" in pois:
elif isSignalStrength or paramFamily == "absxsec" or paramFamily == "normxsec":
keys = pullSummaryMap.keys()
if analysis == "helicity":
keys_l = list(k for k in keys if 'left' in k)
keys_r = list(k for k in keys if 'right' in k)
norms_l = sorted(keys_l, key=lambda k: int(k.split('_')[-1]), reverse=False)
norms_r = sorted(keys_r, key=lambda k: int(k.split('_')[-1]), reverse=True)
sortedpulls = norms_r + norms_l
else:
# differential xsection: get ieta, ipt index and use them as keys to sort
tmpkeys = [x for x in keys if "outliers" not in x]
sortedpulls = sorted(tmpkeys, key = lambda x: get_ieta_ipt_from_process_name(x))
for x in keys:
if "outliers" in x: sortedpulls.append(x)
else:
sortedpulls = sorted(pullSummaryMap.keys(), key=lambda k: k[0]) # alphabetic order
if len(sortedpulls)==0: break
maxErr = 0;
for name in sortedpulls:
if pi>nThisPulls: break
pull = pullSummaryMap[name]
pull_rms.SetBinContent(3*pi+0,pull[0])
pull_rms.SetBinError(3*pi+0,pull[1])
pull_effsigma.SetBinContent(3*pi+1,pull[2])
pull_effsigma.SetBinError(3*pi+1,pull[3])
if useMedian:
pull_median.SetBinContent(3*pi+1,mean_p)
pull_median.SetBinError(3*pi+1,0)
newname = name
# if isSignalStrength:
# newname = "_".join(name.split("_")[:6])
# elif paramFamily == "absxsec":
# newname = "_".join(name.split("_")[:6]) + "_xsec"
# elif paramFamily == "normxsec":
# newname = "_".join(name.split("_")[:6]) + "_xsecNorm"
if isSignalStrength or paramFamily == "absxsec" or paramFamily == "normxsec":
newname = niceName(name)
pull_rms.GetXaxis().SetBinLabel(3*pi,newname)
pull_rms.GetXaxis().SetTickSize(0.)
max2=max(pull[1],pull[3])
if max2>maxErr: maxErr=max2
del pullSummaryMap[name]
pi += 1
nRemainingPulls -= 1
#pull_rms .SetMarkerColor(46); pull_rms .SetLineColor(45); pull_rms .SetLineWidth(2); pull_rms .SetMarkerSize(1.0); pull_rms .SetMarkerStyle(21);
#pull_effsigma .SetMarkerColor(21); pull_effsigma .SetLineColor(24); pull_effsigma .SetLineWidth(2); pull_effsigma .SetMarkerSize(1.0); pull_effsigma .SetMarkerStyle(22)
# NmaxChar = 1
# for name in sortedpulls:
# NmaxChar = max(NmaxChar,len(name))
# if NmaxChar >= 8:
# hc.SetBottomMargin(0.2) # hardcoded and not optimized I just looked at a random pull plot with the CMS_We_elescale and QCD scales parameters
pull_rms .SetMarkerColor(ROOT.kRed+2);
pull_rms .SetLineColor(ROOT.kRed); pull_rms .SetLineWidth(3);
pull_rms .SetMarkerSize(1.0); pull_rms .SetMarkerStyle(21);
pull_effsigma .SetMarkerColor(ROOT.kBlack);
pull_effsigma .SetLineColor(ROOT.kGray+2); pull_effsigma .SetLineWidth(3);
pull_effsigma .SetMarkerSize(1.0); pull_effsigma .SetMarkerStyle(22)
pull_median.SetMarkerColor(ROOT.kBlack);
pull_median.SetLineColor(ROOT.kBlack);
pull_median.SetLineWidth(3);
pull_median.SetMarkerSize(1.5);
pull_median.SetMarkerStyle(20)
pull_rms.SetLabelSize(pullLabelSize)
pull_rms.LabelsOption("v")
yrange = 3.
if isSignalStrength and not plotPull:
pull_rms.GetYaxis().SetRangeUser(0.9,1.1)
else:
pull_rms.GetYaxis().SetRangeUser(-yrange,yrange)
if plotPull:
pull_rms.GetYaxis().SetTitle("pull summary (n#sigma)")
else:
pull_rms.GetYaxis().SetTitle("POI summary")
pull_rms.Draw("p")
pull_effsigma.Draw("p same")
line0 = ROOT.TLine(pull_rms.GetXaxis().GetXmin(), 0., pull_rms.GetXaxis().GetXmax(), 0.); line0.SetLineStyle(7)
line0.Draw('same')
leg = ROOT.TLegend(0.60, 0.75, 0.85, 0.95)
leg.SetFillStyle(0)
leg.SetBorderSize(0)
leg.AddEntry(pull_rms,"Gaussian #sigma")
leg.AddEntry(pull_effsigma,"Effective #sigma")
leg.Draw("same")
legMedian = ROOT.TLegend(0.15, 0.85, 0.4, 0.95)
if useMedian:
legMedian.SetFillStyle(0)
legMedian.SetBorderSize(0)
legMedian.AddEntry(pull_median,"Median","PL")
legMedian.Draw("same")
param_group=pois.replace('.*','').replace(',','_')
for ext in ['png', 'pdf']:
hc.SaveAs("{pdir}/pullSummaryToys_{params}_{igroup}_{ch}_{c}.{ext}".format(pdir=plotdir,ch=charge,suffix="_"+suffix,params=param_group,igroup=pullPlots,c=channel,ext=ext))
pullPlots += 1
def niceName(name):
if '_Ybin' in name:
nn = '#mu: ' if '_mu_' in name else 'el: '
if 'plus' in name: nn += 'W+ '
elif 'minus' in name: nn += 'W- '
else: nn += 'W '
if 'left' in name: nn += 'left '
elif 'right' in name: nn += 'right '
elif 'long' in name: nn += 'long '
else: nn += 'unpolarized '
idx = -2 if (name.endswith('mu') or any([x in name for x in ['masked','sumxsec','charge','a0','a4']])) else -1
nn += name.split('_')[idx]
if 'eff_unc' in name:
nn = '#epsilon_{unc}^{'+nn+'}'
return nn
elif '_ieta_' and '_ipt_' in name:
nn = '#mu: ' if '_mu_' in name else 'el: '
nn += 'W+ ' if 'plus' in name else 'W- ' if 'minus' in name else 'W '
ieta,ipt = get_ieta_ipt_from_process_name(name)
nn += "i#eta, ip_{{T}} = {neta}, {npt} ".format(neta=ieta,npt=ipt)
return nn
elif '_ieta_' in name:
nn = '#mu: ' if '_mu_' in name else 'el: '
nn += 'W+ ' if 'plus' in name else 'W- ' if 'minus' in name else 'W '
ieta = int((name.split("_ieta_")[1]).split("_")[0])
nn += "i#eta = {neta}".format(neta=ieta)
return nn
elif '_ipt_' in name:
nn = '#mu: ' if '_mu_' in name else 'el: '
nn += 'W+ ' if 'plus' in name else 'W- ' if 'minus' in name else 'W '
ipt = int((name.split("_ipt_")[1]).split("_")[0])
nn += "i#p_{{T}} = {npt}".format(npt=ipt)
return nn
elif "CMS_" in name:
if "CMS_Wmu" in name:
return name.replace("CMS_Wmu_","")
elif "CMS_We" in name:
return name.replace("CMS_We_","")
else:
return name
elif re.match( "Fakes(Eta|Pt)Uncorrelated.*",name):
num = re.findall(r'\d+', name) # get number
pfx = name.split(num[0])[1] # split on number and read what's on the right
leptonCharge = ""
if len(pfx):
leptonCharge = "{lep}{chs}".format(lep="#mu" if "mu" in pfx else "e", chs = "+" if "plus" in pfx else "-" if "minus" in pfx else "")
return "Fakes {var}-uncorr.{n} {lepCh}".format(var="#eta" if "FakesEta" in name else "p_{T}", n=num[0], lepCh=leptonCharge)
elif re.match(".*EffStat\d+.*",name):
num = re.findall(r'\d+', name) # get number (there will be two of them, need the second)
pfx = name.split("EffStat"+str(num[1]))[1] # split on second number and read what's on the right
leptonCharge = ""
if len(pfx):
leptonCharge = "{lep}{chs}".format(lep="#mu" if "mu" in pfx else "e", chs = "+" if "plus" in pfx else "-" if "minus" in pfx else "")
return "Eff. uncorr. {n1}-{n2} {lepCh}".format(n1=num[0],n2=num[1],lepCh=leptonCharge)
else:
return name
# to help sorting with helicity
# if using more helicity and Y bins, sort by hel,Ybin
helSorted = { "left" : 1, "right" : 2, "long" : 3}
chargeSorted = { "Wplus" : 1, "Wminus" : 2}
## sort the floatParams. alphabetically, except for pdfs, which are sorted by number
## for mu* QCD scales, distinguish among muR and muRXX with XX in 1-10
# why is this commented? Isn't it nice that different charge and polarizations are grouped together?
# see old example here:
# http://mciprian.web.cern.ch/mciprian/wmass/13TeV/helicityAnalysis/electron/fromEmanuele/13-12-18/fitresults_poim1_exp1_bbb1/subMatrix/smallCorrelation_Wplusminus_leftrightYbin_2-5.png
#params = sorted(params, key= lambda x: (int(chargeSorted[x.split('_')[0]]),int(helSorted[x.split('_')[1]]),int(x.split('_')[-1])) if '_Ybin_' in x else 0)
# one might want to invert the order of charge and helicity for the sorting
params = sorted(params, key= lambda x: utilities.getNFromString(x) if '_Ybin_' in x else 0)
params = sorted(params, key= lambda x: get_ieta_ipt_from_process_name(x) if ('_ieta_' in x and '_ipt_' in x) else 0)
params = sorted(params, key= lambda x: int(x.replace('pdf','')) if 'pdf' in x else 0)
params = sorted(params, key= lambda x: int(x.replace('muRmuF','')) if ('muRmuF' in x and x != "muRmuF") else 0)
params = sorted(params, key= lambda x: int(x.replace('muR','')) if (''.join([j for j in x if not j.isdigit()]) == 'muR' and x != "muR") else 0)
params = sorted(params, key= lambda x: int(x.replace('muF','')) if (''.join([j for j in x if not j.isdigit()]) == 'muF' and x != "muF") else 0)
params = sorted(params, key= lambda x: utilities.getNFromString(x) if 'EffStat' in x else 0)
params = sorted(params, key= lambda x: utilities.getNFromString(x) if 'FakesEtaUncorrelated' in x else 0)
params = sorted(params, key= lambda x: utilities.getNFromString(x) if 'FakesPtUncorrelated' in x else 0)
# sort by charge if needed
# I think it is useful that different charges are separated in the matrix, it is easier to read it
params = sorted(params, key = lambda x: 0 if "plus" in x else 1 if "minus" in x else 2)
print "sorted params = ", params
c = ROOT.TCanvas("c","",1200,800)
c.SetGridx()