def storeHistogram(var, lumi, cut, histo, filename):
# Plotting Data
# ------------------
bp = BasicPlot(filename, var)
bp.addHistogram(histo, 'E', copy=True)
bp.titles.append(
"#sqrt{{s}} = 13 TeV, #scale[0.5]{{#int}}Ldt = {0:.2f} fb^{{-1}}".
format(lumi / 1000.))
bp.yVariable = Variable("Events")
bp.normalizeByBinWidth = False
bp.showBinWidthY = False
bp.normalized = False
bp.logY = False
bp.yVariable.binning.low = 0 # start y-axis at 0
bp.draw()
bp.saveAsAll(os.path.join(outFolder, bp.title))
# storing histogram in root file
# ------------------------------
from ROOT import TFile, TObject
rootfile = TFile.Open(outFolder + '/' + filename + '.root', 'update')
if rootfile.IsOpen():
print "writing in file {}".format(outFolder + '/' + filename + '.root')
rootfile.cd()
histogram = histo.Clone('h_' + var.name)
histogram.SetTitle('h_' + var.name)
histogram.Write('h_' + var.name, TObject.kOverwrite)
rootfile.Close()
else:
print "Problem opening file {}".format(outFolder + '/' + filename +
'.root')
def PlotAndStore(var,
filename,
lumi,
h_1,
h_2=None,
h_3=None,
drawOneLine=False,
simulation=False,
yLabel="Fake Rate",
yLow=0,
yHigh=None):
bp = BasicPlot(filename, var)
# Add histos
SetHistStyle(h_1, 0)
bp.addHistogram(h_1, 'E', copy=True)
if h_2:
SetHistStyle(h_2, 1)
bp.addHistogram(h_2, 'E', copy=True)
if h_3:
SetHistStyle(h_3, 2)
bp.addHistogram(h_3, 'E', copy=True)
# Draw a line at 1 e.g. for SF plots
if drawOneLine:
h_tmp = var.createHistogram('')
for i_bin in range(h_tmp.GetNbinsX() + 2):
h_tmp.SetBinContent(i_bin, 1)
h_tmp.SetBinError(i_bin, 0)
h_tmp.SetLineColor(kGray)
bp.addHistogram(h_tmp, 'E')
# Some further settings
bp.titles.append(
'#sqrt{{s}} = 13 TeV, #scale[0.5]{{#int}}Ldt = {0:.2f} fb^{{-1}}'.
format(lumi / 1000.))
bp.legendDecorator.textSize = 0.045
if simulation:
bp.titles.append('Simulation')
bp.showBinWidthY = False
bp.logY = False
bp.normalized = False
bp.normalizeByBinWidth = False
bp.yVariable = Variable(yLabel)
bp.yVariable.binning.low = yLow
bp.yVariable.binning.up = yHigh
# Plot and store
bp.draw()
bp.saveAsAll(os.path.join(outFolder, bp.title))
cut=selection,
weightExpression=weight_expression)
sys.stdout.write('.')
sys.stdout.flush()
h_data.Add(h_tmp, -1)
print "read all histograms for {0}".format(var.name)
# calculating scale factor data/MC for the given Variable
h_sf = var.createHistogram('SF')
h_sf.Divide(h_data, h_mc, 1, 1, 'B')
# plotting the scale factor
bp_sf = BasicPlot('SF_' + var.name, var, Variable('Scale Factor'))
bp_sf.titles.append(
'#sqrt{{s}} = 13 TeV, #scale[0.5]{{#int}}Ldt = {0:.2f} fb^{{-1}}, '
.format(lumi / 1000.))
bp_sf.addHistogram(h_sf, 'E', copy=True)
bp_sf.draw()
bp_sf.saveAsAll(os.path.join(outFolder, bp_sf.title))
# storing histogram in root file
from ROOT import TFile, TObject
filename = outFolder + '/SF_' + var.name + '.root'
rootfile = TFile.Open(filename, 'update')
if rootfile.IsOpen():
print "writing in file {}".format(filename)
rootfile.cd()
histogram = h_sf.Clone(var.name)
histogram.SetTitle(var.name)
histogram.Write(var.name, TObject.kOverwrite)
rootfile.Close()
else:
print "Problem opening file {}".format(filename)
for i_set in range(n_sets): # data or mc plot ?
if i_type == 2: # SF plots have a different naming convention
bp = BasicPlot( 'FakeRate_'+var.name+name_type[i_type]+sel_p_suffix[p], var )
else:
bp = BasicPlot( 'FakeRate_'+var.name+name_type[i_type]+sel_p_suffix[p]+name_set[i_set], var )
bp.showBinWidthY = False
if name_type[i_type] == '_SF': # Draw a line at 1 for SF plots
h_tmp = var.createHistogram( '' )
for i_bin in range( h_tmp.GetNbinsX() + 2 ):
h_tmp.SetBinContent( i_bin, 1 )
h_tmp.SetBinError( i_bin, 0 )
h_tmp.SetLineColor( kGray )
bp.addHistogram( h_tmp, 'E' )
for i in range(id_range[i_type]): # Loop over different IDs
histo_set[len(name_set)*i_type+i_set][i].SetLineColor( sel_colors[i] )
histo_set[len(name_set)*i_type+i_set][i].SetMarkerColor( sel_colors[i] )
histo_set[len(name_set)*i_type+i_set][i].SetMarkerStyle( sel_marker[i] )
bp.addHistogram( histo_set[len(name_set)*i_type+i_set][i], 'E', copy=True )
bp.titles.append( '#sqrt{{s}} = 13 TeV, #scale[0.5]{{#int}}Ldt = {0:.2f} fb^{{-1}}'.format(lumi/1000.) )
bp.legendDecorator.textSize = 0.045
bp.normalizeByBinWidth = normByWidth[i_type]
bp.yVariable = Variable( yLabel[i_type] )
bp.logY = logScale[i_type]
bp.normalized = False
if i_type == 1 or i_type == 2: # For FR and SF plots start y-axis at 0
bp.yVariable.binning.low = 0
bp.draw()
bp.saveAsAll( os.path.join( "plots/", bp.title ) )
# --------------------
print 'everything is done!'
def TemplateFit(filename,
data,
mc_q,
mc_g,
w_q=None,
w_g=None,
pull_widths=[1, 1],
doPlots=True):
n_mc = 2
# Define MC samples
mc = TObjArray(n_mc)
mc.Add(mc_q)
mc.Add(mc_g)
# Names
mc_type = []
mc_type.append('Quarks')
mc_type.append('Gluons')
# Perform Fit
fit = TFractionFitter(data, mc, "q")
if w_q and w_g:
fit.SetWeight(0, w_q)
fit.SetWeight(1, w_g)
fit.Constrain(0, 0.0, 1.0) #quarks
fit.Constrain(1, 0.0, 1.0) #gluons
fit.Fit()
# Printing results
val_q = ROOT.Double()
err_q = ROOT.Double()
val_g = ROOT.Double()
err_g = ROOT.Double()
fit.GetResult(0, val_q, err_q)
fit.GetResult(1, val_g, err_g)
err_q = err_q * pull_widths[
0] # correcting the error for the width of the pull plot
err_g = err_g * pull_widths[
1] # correcting the error for the width of the pull plot
for i in range(n_mc):
val = ROOT.Double()
err = ROOT.Double()
fit.GetResult(i, val, err)
out_file.write("fit result {}: {}, {}\n".format(mc_type[i], val, err))
out_file.write("X2 = {}\n".format(fit.GetChisquare()))
out_file.write("NDF = {}\n".format(fit.GetNDF()))
red_chi = fit.GetChisquare() / fit.GetNDF()
out_file.write("red. X2 = {}\n".format(red_chi))
out_file.write("\n")
if doPlots:
print "storing plot as {}".format(filename)
# Preparing Histograms
h_fit = fit.GetPlot()
h_q = fit.GetMCPrediction(0)
h_g = fit.GetMCPrediction(1)
h_fit.SetTitle("Fit")
h_q.SetTitle("Quarks")
h_g.SetTitle("Gluons")
h_fit.SetLineColor(kGreen)
h_q.SetLineColor(kRed)
h_g.SetLineColor(kBlue)
h_q.SetMarkerColor(kRed)
h_g.SetMarkerColor(kBlue)
# Scaling to the fit result
int_fit = h_fit.Integral()
int_q = h_q.Integral()
int_g = h_g.Integral()
h_q.Sumw2()
h_g.Sumw2()
h_q.Scale(val_q * int_fit / int_q)
h_g.Scale(val_g * int_fit / int_g)
# Checking fit vs. templates
h_check = h_fit.Clone('Fit - Templates Check')
h_check.Add(h_q, -1)
h_check.Add(h_g, -1)
check_maxBin = h_check.GetMaximumBin()
out_file.write(
"Maximal bin of fit minus templates divided by content of that bin in fit: {} / {}\n\n"
.format(h_fit.GetBinContent(check_maxBin),
h_check.GetBinContent(check_maxBin)))
# Plotting results
bp = BasicPlot(filename, var_tau0_width)
bp.addHistogram(h_q, 'E', copy=True)
bp.addHistogram(h_g, 'E', copy=True)
bp.addHistogram(h_fit, 'HIST', copy=True)
bp.addHistogram(data, 'E', copy=True)
bp.titles.append(
'Quarks: {:.3} #pm {:.2}, Gluons: {:.3} #pm {:.2}'.format(
val_q, err_q, val_g, err_g))
bp.titles.append(
'#sqrt{{s}} = 13 TeV, #scale[0.5]{{#int}}Ldt = {0:.2f} fb^{{-1}}'.
format(lumi / 1000.))
bp.titles.append('#chi^{{2}}/ndf: {:.3}'.format(red_chi))
bp.yVariable = Variable("Events")
bp.legendDecorator.textSize = 0.045
bp.normalizeByBinWidth = False
bp.showBinWidthY = False
bp.normalized = False
bp.logY = False
bp.yVariable.binning.low = 0 # start y-axis at 0
bp.draw()
bp.saveAsAll(os.path.join(outFolder, bp.title))
return fit
def PullPlot(filename,
h_data,
h_quarks,
h_gluons,
w_quarks=None,
w_gluons=None,
pull_widths=[1, 1]):
# Creating a pull-plot
# --------------------
ratio_q = []
error_q = []
ratio_g = []
error_g = []
n_runs = 10000
# getting fit results for toy experiments
for i in range(1, n_runs + 1):
if i == 1: print " starting loop"
h_tmp = var.createHistogram(str(i))
h_tmp.FillRandom(h_data)
filename_tmp = filename + '_' + str(i)
out_file.write(str(i) + ":\n")
fit = TemplateFit(filename,
h_tmp,
h_quarks,
h_gluons,
w_quarks,
w_gluons,
pull_widths,
doPlots=False)
val_q = ROOT.Double()
err_q = ROOT.Double()
val_g = ROOT.Double()
err_g = ROOT.Double()
fit.GetResult(0, val_q, err_q)
fit.GetResult(1, val_g, err_g)
del fit
err_q = err_q * pull_widths[
0] # correcting the error for the width of the pull plot
err_g = err_g * pull_widths[
1] # correcting the error for the width of the pull plot
ratio_q.append(val_q)
error_q.append(err_q)
ratio_g.append(val_g)
error_g.append(err_g)
if i % (n_runs / 10) == 0 or i == 1 or i == n_runs:
print " {}\tQ: {} +- {}\tG: {} +- {}".format(
i, val_q, err_q, val_g, err_g)
# creating pull histogram for quarks
h_pull = var_pull.createHistogram()
ratio_q_mean = sum(ratio_q) / len(ratio_q)
for i in range(n_runs):
tmp = (ratio_q[i] - ratio_q_mean) / error_q[i]
h_pull.Fill(tmp)
fit_result = h_pull.Fit('gaus', 'QS')
f_gaus = h_pull.GetFunction('gaus')
v_q_mu = fit_result.Parameter(1)
v_q_mu_err = fit_result.ParError(1)
v_q_sigma = fit_result.Parameter(2)
v_q_sigma_err = fit_result.ParError(2)
chi = fit_result.Chi2()
ndf = fit_result.Ndf()
#red_chi = chi / ndf
filename_tmp = filename + '_PullPlot_q'
bp = BasicPlot(filename_tmp, var_pull)
bp.titles.append(
'Fit Results: #mu = {:05.3f}#pm{:05.3f}, #sigma = {:05.3f}#pm{:05.3f}'.
format(v_q_mu, v_q_mu_err, v_q_sigma, v_q_sigma_err))
bp.titles.append('#chi^{{2}}/ndf: {:.3}/{}'.format(chi, ndf))
bp.addHistogram(h_pull, 'E', copy=True)
bp.yVariable = Variable("Events")
bp.normalizeByBinWidth = False
bp.showBinWidthY = False
bp.normalized = False
bp.logY = False
bp.yVariable.binning.low = 0 # start y-axis at 0
bp.draw()
bp.saveAsAll(os.path.join(outFolder, bp.title))
del h_pull
del bp
# creating pull histogram for gluons
h_pull = var_pull.createHistogram()
ratio_g_mean = sum(ratio_g) / len(ratio_g)
for i in range(n_runs):
tmp = (ratio_g[i] - ratio_g_mean) / error_g[i]
h_pull.Fill(tmp)
fit_result = h_pull.Fit('gaus', 'QS')
f_gaus = h_pull.GetFunction('gaus')
v_g_mu = fit_result.Parameter(1)
v_g_mu_err = fit_result.ParError(1)
v_g_sigma = fit_result.Parameter(2)
v_g_sigma_err = fit_result.ParError(2)
chi = fit_result.Chi2()
ndf = fit_result.Ndf()
#red_chi = chi / ndf
filename_tmp = filename + '_PullPlot_g'
bp = BasicPlot(filename_tmp, var_pull)
bp.titles.append(
'Fit Results: #mu = {:05.3f}#pm{:05.3f}, #sigma = {:05.3f}#pm{:05.3f}'.
format(v_g_mu, v_g_mu_err, v_g_sigma, v_g_sigma_err))
bp.titles.append('#chi^{{2}}/ndf: {:.3}/{}'.format(chi, ndf))
bp.addHistogram(h_pull, 'E', copy=True)
bp.yVariable = Variable("Events")
bp.normalizeByBinWidth = False
bp.showBinWidthY = False
bp.normalized = False
bp.logY = False
bp.yVariable.binning.low = 0 # start y-axis at 0
bp.draw()
bp.saveAsAll(os.path.join(outFolder, bp.title))
return [v_q_sigma, v_g_sigma]