def mc_stack(
hlist,
hs_syst,
systematics,
colors="auto"
):
if colors=="auto":
coloriter = iter(plt.cm.jet(np.linspace(0,1,len(hlist))))
for h in hlist:
h.color = next(coloriter)
elif isinstance(colors, list) and len(colors) == len(hlist):
for h, c in zip(hlist, colors):
h.color = c
for h in hlist:
h.fillstyle = "solid"
#FIXME: Temporary workaround for failed fill, only works when hatch is specified
stack = hist(hlist, stacked=True, hatch=".", lw=2)
htot = sum(hlist)
htot.color="black"
htot_u = htot.Clone()
htot_d = htot.Clone()
for i in range(1, htot.nbins()+1):
htot_u.set_bin_content(i, htot.get_bin_content(i) + htot.get_bin_error(i))
htot_d.set_bin_content(i, htot.get_bin_content(i) - htot.get_bin_error(i))
htot_u.color="black"
htot_d.color="black"
fill_between(htot_u, htot_d,
color="black", hatch="////////",
alpha=1.0, linewidth=0, facecolor="none", edgecolor="black", zorder=10,
)
#add systematic uncertainties
hstat = htot_u - htot_d
errs = np.array([y for y in hstat.y()])
errs = np.abs(errs)
htot_usyst = htot.Clone()
htot_dsyst = htot.Clone()
for systUp, systDown in systematics:
errs_syst_up = np.array([y for y in sum(hs_syst[systUp].values()).y()])
errs_syst_down = np.array([y for y in sum(hs_syst[systDown].values()).y()])
errs_syst = np.abs(errs_syst_up - errs_syst_down)
errs = np.power(errs, 2) + np.power(errs_syst, 2)
errs = np.sqrt(errs)
for i in range(len(errs)):
htot_usyst.SetBinContent(i+1, htot_usyst.GetBinContent(i+1) + errs[i]/2)
htot_dsyst.SetBinContent(i+1, htot_dsyst.GetBinContent(i+1) - errs[i]/2)
fill_between(htot_usyst, htot_dsyst,
color="gray", hatch=r"\\\\",
alpha=1.0, linewidth=0, facecolor="none", edgecolor="gray", zorder=10,
)
return {"hists":stack, "tot":htot, "tot_u":htot_u, "tot_d":htot_d, "tot_usyst":htot_usyst, "tot_dsyst":htot_dsyst}
def draw(self, axes,
facecolor = '0.75', # grey
alpha = 0.5,
hatch = '/',
zorder = 999,):
rplt.fill_between( self.__upper,
self.__lower, axes, facecolor = facecolor,
alpha = alpha, hatch = hatch,
zorder = zorder )
def draw(
self,
axes,
facecolor='0.75', # grey
alpha=0.5,
hatch='/',
zorder=999,
):
rplt.fill_between(self.__upper,
self.__lower,
axes,
facecolor=facecolor,
alpha=alpha,
hatch=hatch,
zorder=zorder)
def plot_systematic_uncertainties(systematic_uncertainties,
bin_edges,
variable,
output_folder,
subcategories=[],
subname='',
plot_largest=False,
plot_foreground=None):
'''
Plot the systematic uncertainties
'''
if not subcategories: subcategories = systematic_uncertainties.keys()
x_limits = [bin_edges[0], bin_edges[-1]]
# y_limits = [-0.6,0.6]
y_limits = [0, 0.4]
fig_syst = plt.figure(figsize=(20, 16), dpi=400, facecolor='white')
ax_syst = fig_syst.add_subplot(1, 1, 1)
ax_syst.minorticks_on()
ax_syst.xaxis.labelpad = 12
ax_syst.yaxis.labelpad = 12
error_hists_up = {}
error_hists_down = {}
stat_hist = None
for syst, vals in systematic_uncertainties.iteritems():
if syst == 'central':
n = len(systematic_uncertainties[syst])
continue
elif syst == 'statistical':
stat_hist_up = values_and_errors_to_hist(vals, [], bin_edges)
stat_hist_down = values_and_errors_to_hist(-vals, [], bin_edges)
elif syst == 'systematic':
syst_hist_up = values_and_errors_to_hist(vals, [], bin_edges)
syst_hist_down = values_and_errors_to_hist(-vals, [], bin_edges)
elif syst in subcategories:
error_hists_up[syst] = values_and_errors_to_hist(
vals, [], bin_edges)
error_hists_down[syst] = values_and_errors_to_hist(
-vals, [], bin_edges)
else:
continue
if plot_largest:
largest_syst = []
for bin_i in range(n):
high = []
for syst, vals in systematic_uncertainties.iteritems():
if syst == 'central': continue
if syst == 'statistical': continue
if syst == 'systematic': continue
high.append([syst, vals[bin_i]])
high = sorted(high, key=itemgetter(1), reverse=True)
# Retrieve highest systematics
if high[0][0] not in largest_syst: largest_syst.append(high[0][0])
elif high[1][0] not in largest_syst:
largest_syst.append(high[1][0])
else:
continue
rplt.fill_between(syst_hist_up,
syst_hist_down,
color='gold',
label='Syst.')
rplt.fill_between(stat_hist_down,
stat_hist_up,
color='0.75',
label='Stat.')
plt.tick_params(**CMS.axis_label_major)
plt.tick_params(**CMS.axis_label_minor)
colours = [
'red', 'blue', 'green', 'chartreuse', 'indigo', 'magenta',
'darkmagenta', 'hotpink', 'cyan', 'darkred', 'darkgoldenrod',
'mediumvioletred', 'mediumspringgreen', 'darkgoldenrod', 'slategray',
'dodgerblue', 'cadetblue', 'darkblue', 'seagreen', 'deeppink',
'deepskyblue'
] * 2
if len(colours) < len(error_hists_up.keys()):
print '---> Need to add more colours!!!'
for error_hists in [error_hists_up, error_hists_down]:
for i, source, in enumerate(error_hists.keys()):
hist = error_hists[source]
hist.linewidth = 4
hist.color = colours[i]
if plot_largest:
if source not in largest_syst:
hist.linestyle = 'dashed'
hist.alpha = 0.4
hist.linewidth = 2
# Only label systematic once
if plot_foreground and plot_foreground in error_hists.keys():
source = ''
if error_hists == error_hists_up:
rplt.hist(hist, stacked=False, label=source)
else:
rplt.hist(hist, stacked=False, label='')
if plot_foreground and plot_foreground in error_hists.keys():
hist = error_hists[plot_foreground]
hist.color = 'black'
rplt.hist(hist, stacked=False, label=source)
leg = plt.legend(loc='upper right', prop={'size': 25}, ncol=3)
# leg = plt.legend(loc='upper right',prop={'size':20},ncol=4)
leg.draw_frame(False)
x_title = variables_NonLatex[variable]
if variable in ['HT', 'MET', 'WPT', 'ST', 'lepton_pt']:
x_title += ' [GeV]'
ax_syst.set_xlim(x_limits)
ax_syst.set_ylim(y_limits)
plt.xlabel(x_title, CMS.x_axis_title)
plt.ylabel('Relative Uncertainty', CMS.y_axis_title)
template = '%.1f fb$^{-1}$ (%d TeV)'
label = template % (measurement_config.new_luminosity / 1000.,
measurement_config.centre_of_mass_energy)
plt.title(label, loc='right', **CMS.title)
logo_location = (0.05, 0.98)
prelim_location = (0.05, 0.92)
channel_location = (0.05, 0.86)
# plt.text(logo_location[0], logo_location[1],
# "CMS",
# transform=ax_syst.transAxes,
# fontsize=42,
# verticalalignment='top',
# horizontalalignment='left'
# )
# # preliminary
# plt.text(prelim_location[0], prelim_location[1],
# r"\emph{Preliminary}",
# transform=ax_syst.transAxes,
# fontsize=42,
# verticalalignment='top',
# horizontalalignment='left'
# )
# # channel text
# plt.text(channel_location[0], channel_location[1],
# r"\emph{%s}" % channel,
# transform=ax_syst.transAxes,
# fontsize=40,
# verticalalignment='top',
# horizontalalignment='left'
# )
plt.tight_layout()
file_template = output_folder + '{var}_systematics_{com}TeV'.format(
var=variable,
com=measurement_config.centre_of_mass_energy,
)
if subname: file_template = file_template + '_' + subname
file_template += '.pdf'
fig_syst.savefig(file_template)
print "Written plots to {f}".format(f=file_template)
# plt.show()
return
def plot_systematic_uncertainties(systematic_uncertainties, bin_edges, variable, output_folder, subcategories = [], subname = '', plot_largest = False):
'''
Plot the systematic uncertainties
'''
print subcategories
if not subcategories: subcategories = systematic_uncertainties.keys()
x_limits = [bin_edges[0], bin_edges[-1]]
# y_limits = [-0.6,0.6]
y_limits = [0,0.4]
fig_syst = plt.figure( figsize = ( 20, 16 ), dpi = 400, facecolor = 'white' )
ax_syst = fig_syst.add_subplot(1, 1, 1)
ax_syst.minorticks_on()
ax_syst.xaxis.labelpad = 12
ax_syst.yaxis.labelpad = 12
error_hists_up = {}
error_hists_down = {}
stat_hist = None
for syst, vals in systematic_uncertainties.iteritems():
if syst == 'central':
n = len(systematic_uncertainties[syst])
continue
elif syst == 'statistical':
stat_hist_up = values_and_errors_to_hist( vals, [], bin_edges )
stat_hist_down = values_and_errors_to_hist( -vals, [], bin_edges )
elif syst == 'systematic':
syst_hist_up = values_and_errors_to_hist( vals, [], bin_edges )
syst_hist_down = values_and_errors_to_hist( -vals, [], bin_edges )
elif syst in subcategories:
error_hists_up[syst] = values_and_errors_to_hist( vals, [], bin_edges )
error_hists_down[syst] = values_and_errors_to_hist( -vals, [], bin_edges )
else: continue
if plot_largest:
largest_syst = []
for bin_i in range( n ):
high = []
for syst, vals in systematic_uncertainties.iteritems():
if syst == 'central': continue
if syst == 'statistical': continue
if syst == 'systematic': continue
high.append([syst,vals[bin_i]])
high = sorted(high, key = itemgetter(1), reverse=True)
# Retrieve highest systematics
if high[0][0] not in largest_syst: largest_syst.append(high[0][0])
elif high[1][0] not in largest_syst: largest_syst.append(high[1][0])
else: continue
rplt.fill_between( syst_hist_up, syst_hist_down, color = 'yellow', label='Syst.' )
rplt.fill_between( stat_hist_down, stat_hist_up, color = 'grey', label='Stat.' )
plt.tick_params( **CMS.axis_label_major )
plt.tick_params( **CMS.axis_label_minor )
colours = ['red', 'blue', 'green', 'chartreuse', 'indigo', 'magenta', 'darkmagenta', 'hotpink', 'cyan', 'darkred', 'darkgoldenrod', 'mediumvioletred', 'mediumspringgreen', 'gold', 'darkgoldenrod', 'slategray', 'dodgerblue', 'cadetblue', 'darkblue', 'seagreen', 'deeppink', 'deepskyblue' ]
# if len(colours) < len(error_hists.keys()):
# print '---> Need to add more colours!!!'
for error_hists in [error_hists_up, error_hists_down]:
for i, source, in enumerate(error_hists.keys()):
hist = error_hists[source]
hist.linewidth = 4
hist.color = colours[i]
if plot_largest:
if source not in largest_syst:
hist.linestyle = 'dashed'
hist.alpha = 0.4
hist.linewidth = 2
# Only label systematic once
if error_hists == error_hists_up:
rplt.hist( hist, stacked=False, label = source )
else:
rplt.hist( hist, stacked=False, label = '' )
leg = plt.legend(loc='upper right',prop={'size':25},ncol=3)
# leg = plt.legend(loc='upper right',prop={'size':20},ncol=4)
leg.draw_frame(False)
x_title = variables_NonLatex[variable]
if variable in ['HT', 'MET', 'WPT', 'ST', 'lepton_pt']:
x_title += ' [GeV]'
ax_syst.set_xlim( x_limits )
ax_syst.set_ylim( y_limits )
plt.xlabel( x_title, CMS.x_axis_title )
plt.ylabel( 'Relative Uncertainty', CMS.y_axis_title)
template = '%.1f fb$^{-1}$ (%d TeV)'
label = template % ( measurement_config.new_luminosity/1000, measurement_config.centre_of_mass_energy)
plt.title( label,loc='right', **CMS.title )
logo_location = (0.05, 0.98)
prelim_location = (0.05, 0.92)
channel_location = ( 0.05, 0.86)
# plt.text(logo_location[0], logo_location[1],
# "CMS",
# transform=ax_syst.transAxes,
# fontsize=42,
# verticalalignment='top',
# horizontalalignment='left'
# )
# # preliminary
# plt.text(prelim_location[0], prelim_location[1],
# r"\emph{Preliminary}",
# transform=ax_syst.transAxes,
# fontsize=42,
# verticalalignment='top',
# horizontalalignment='left'
# )
# # channel text
# plt.text(channel_location[0], channel_location[1],
# r"\emph{%s}" % channel,
# transform=ax_syst.transAxes,
# fontsize=40,
# verticalalignment='top',
# horizontalalignment='left'
# )
plt.tight_layout()
file_template = output_folder + '{var}_systematics_{com}TeV'.format(
var = variable,
com = measurement_config.centre_of_mass_energy,
)
if subname: file_template = file_template + '_' + subname
file_template += '.pdf'
fig_syst.savefig(file_template)
print "Written plots to {f}".format(f = file_template)
# plt.show()
return
def make_data_mc_comparison_plot(
histograms=[],
histogram_lables=[],
histogram_colors=[],
histogram_properties=Histogram_properties(),
data_index=0,
save_folder='plots/',
save_as=['pdf'],
normalise=False,
show_ratio=False,
show_stat_errors_on_mc=False,
draw_vertical_line=0,
systematics_for_ratio=None,
systematics_for_plot=None,
histograms_to_compare=None,
):
'''
systematics_for_plot takes the same input as systematics_for_ratio. There may be some repition with reagrds to
mc_error and mc_relative_errors, but these only deal with a flat error up and down.
'''
save_folder = check_save_folder(save_folder)
# make copies in order not to mess with existing histograms
histograms_ = deepcopy(histograms)
stack = HistStack()
add_mc = stack.Add
for index, histogram in enumerate(histograms_):
label = histogram_lables[index]
color = histogram_colors[index]
histogram.SetTitle(label)
if normalise:
histogram.Sumw2()
if not index == data_index:
histogram.fillstyle = 'solid'
histogram.fillcolor = color
histogram.legendstyle = 'F'
add_mc(histogram)
data = histograms_[data_index]
data.SetMarkerSize(CMS.data_marker_size)
if normalise:
n_events_data = data.Integral()
n_events_mc = stack.Integral()
data.Scale(1 / n_events_data)
stack.Scale(1 / n_events_mc)
# plot with matplotlib
plt.figure(figsize=CMS.figsize, dpi=CMS.dpi, facecolor=CMS.facecolor)
axes = None
if show_ratio:
ratio = data.Clone('ratio')
sumHists = sum(stack.GetHists())
for bin_i in range(1, sumHists.GetNbinsX()):
sumHists.SetBinError(bin_i, 0)
ratio.Divide(sum(stack.GetHists()))
ratio.SetMarkerSize(3)
gs = gridspec.GridSpec(2, 1, height_ratios=[5, 1])
axes = plt.subplot(gs[0])
else:
axes = plt.axes()
if histogram_properties.set_log_y:
axes.set_yscale('log', nonposy="clip")
axes.set_ylim(ymin=1e-2)
if systematics_for_plot != None:
plusErrors = [x + 1 for x in systematics_for_plot]
minusErrors = [1 - x for x in systematics_for_plot]
stack_lower = sum(stack.GetHists())
stack_upper = stack_lower.Clone('upper')
for bin_i in range(1, stack_lower.GetNbinsX() + 1):
central_value = stack_lower.GetBinContent(bin_i)
error_upper_bound = plusErrors[bin_i - 1] * central_value
error_lower_bound = minusErrors[bin_i - 1] * central_value
stack_upper.SetBinContent(bin_i, error_upper_bound)
stack_lower.SetBinContent(bin_i, error_lower_bound)
rplt.fill_between(
stack_lower,
stack_upper,
axes,
hatch='//',
# facecolor = 'Black',
facecolor='None',
edgecolor='Grey',
alpha=1.,
linewidth=0.,
zorder=len(histograms_) + 1)
mc_error = histogram_properties.mc_error
mc_relative_errors = histogram_properties.mc_relative_errors
if mc_relative_errors:
stack_lower = sum(stack.GetHists())
stack_upper = stack_lower.Clone('upper')
for bin_i in range(1, stack_lower.GetNbinsX()):
central_value = stack_lower.GetBinContent(bin_i)
relative_error = mc_relative_errors[bin_i - 1]
error_upper_bound = central_value * (1 + relative_error)
error_lower_bound = central_value * (1 - relative_error)
stack_lower.SetBinContent(bin_i, error_upper_bound)
stack_upper.SetBinContent(bin_i, error_lower_bound)
rplt.fill_between(stack_upper,
stack_lower,
axes,
facecolor='0.75',
alpha=0.5,
hatch='/',
zorder=len(histograms_) + 1)
else:
if mc_error > 0:
stack_lower = sum(stack.GetHists())
stack_upper = stack_lower.Clone('upper')
stack_lower.Scale(1 - mc_error)
stack_upper.Scale(1 + mc_error)
rplt.fill_between(stack_upper,
stack_lower,
axes,
facecolor='0.75',
alpha=0.5,
hatch='/',
zorder=len(histograms_) + 1)
if not mc_error > 0 and show_stat_errors_on_mc:
stack_lower = sum(stack.GetHists())
mc_errors = list(stack_lower.yerravg())
stack_upper = stack_lower.Clone('upper')
for bin_i in range(1, stack_lower.GetNbinsX()):
central_value = stack_lower.GetBinContent(bin_i)
error = mc_errors[bin_i - 1]
error_upper_bound = central_value + error
error_lower_bound = central_value - error
stack_lower.SetBinContent(bin_i, error_lower_bound)
stack_upper.SetBinContent(bin_i, error_upper_bound)
rplt.fill_between(stack_upper,
stack_lower,
axes,
facecolor='0.75',
alpha=0.5,
hatch='/',
zorder=len(histograms_) + 1)
# a comment on zorder: the MC stack should be always at the very back (z = 1),
# then the MC error (z = len(histograms_) + 1) and finally the data
# (z = len(histograms_) + 2)
rplt.hist(stack, stacked=True, axes=axes, zorder=1)
rplt.errorbar(data,
emptybins=histogram_properties.emptybins,
axes=axes,
xerr=histogram_properties.xerr,
elinewidth=2,
capsize=10,
capthick=2,
zorder=len(histograms_) + 2)
if histograms_to_compare:
h_compare = {}
for h, l, c in zip(histograms_to_compare['hists'],
histograms_to_compare['labels'],
histograms_to_compare['colours']):
for histogram in histograms_:
if histogram.GetTitle() not in [
histograms_to_compare['to_replace'], 'data'
]:
h += histogram
h_compare[l] = [h, c]
rplt.step(
h,
axes=axes,
label=l,
color=c,
linewidth=4,
)
# put legend into the correct order (data is always first!)
handles, labels = axes.get_legend_handles_labels()
data_label_index = labels.index('data')
data_handle = handles[data_label_index]
labels.remove('data')
handles.remove(data_handle)
labels.insert(0, 'data')
handles.insert(0, data_handle)
if mc_error > 0 or (not mc_error > 0 and show_stat_errors_on_mc):
p1 = Rectangle((0, 0), 1, 1, fc="0.75", alpha=0.5, hatch='/')
handles.append(p1)
labels.append(histogram_properties.mc_errors_label)
l1 = axes.legend(handles,
labels,
numpoints=1,
frameon=histogram_properties.legend_color,
bbox_to_anchor=histogram_properties.legend_location,
bbox_transform=plt.gcf().transFigure,
prop=CMS.legend_properties,
ncol=histogram_properties.legend_columns)
l1.set_zorder(102)
set_labels(plt,
histogram_properties,
show_x_label=not show_ratio,
axes=axes)
x_limits = histogram_properties.x_limits
y_limits = histogram_properties.y_limits
if len(x_limits) >= 2:
axes.set_xlim(xmin=x_limits[0], xmax=x_limits[-1])
if len(y_limits) >= 2:
axes.set_ylim(ymin=y_limits[0], ymax=y_limits[-1])
else:
y_max = get_best_max_y(
histograms_, x_limits=x_limits) * histogram_properties.y_max_scale
print("Chosen limits : ", 0, y_max)
axes.set_ylim(ymin=0, ymax=y_max)
if histogram_properties.set_log_y:
if not len(y_limits) == 2: # if not user set y-limits, set default
axes.set_ylim(ymin=1e-1)
#draw a red vertical line if needed:
if draw_vertical_line != 0:
plt.axvline(x=draw_vertical_line, color='red', linewidth=3)
if show_ratio:
plt.setp(axes.get_xticklabels(), visible=False)
ax1 = plt.subplot(gs[1])
ax1.minorticks_on()
ax1.grid(True, 'major', linewidth=1)
ax1.axhline(y=1, linewidth=1)
set_labels(plt,
histogram_properties,
show_x_label=True,
show_title=False)
plt.ylabel(r'$\frac{\mathrm{data}}{\mathrm{pred.}}$', CMS.y_axis_title)
ax1.yaxis.set_label_coords(-0.115, 0.8)
rplt.errorbar(ratio,
emptybins=histogram_properties.emptybins,
axes=ax1,
xerr=histogram_properties.xerr,
elinewidth=1.5,
capsize=5,
capthick=1.5)
if histograms_to_compare:
for l, h in h_compare.iteritems():
r = data.Clone(l).Divide(h[0])
rplt.step(
r,
axes=ax1,
label='',
colour=h[1],
linewidth=2,
)
if len(x_limits) >= 2:
ax1.set_xlim(xmin=x_limits[0], xmax=x_limits[-1])
if len(histogram_properties.ratio_y_limits) >= 2:
ax1.set_ylim(ymin=histogram_properties.ratio_y_limits[0],
ymax=histogram_properties.ratio_y_limits[-1])
# dynamic tick placement
adjust_ratio_ticks(ax1.yaxis,
n_ticks=3,
y_limits=histogram_properties.ratio_y_limits)
if histogram_properties.integerXVariable:
ax1.tick_params(axis='x', which='minor', bottom='off', top='off')
if systematics_for_ratio != None:
plusErrors = [x + 1 for x in systematics_for_ratio]
minusErrors = [1 - x for x in systematics_for_ratio]
ratioPlusError = ratio.Clone('plus')
ratioMinusError = ratio.Clone('minus')
for bin_i in range(1, ratioPlusError.GetNbinsX() + 1):
ratioPlusError.SetBinContent(bin_i, plusErrors[bin_i - 1])
ratioMinusError.SetBinContent(bin_i, minusErrors[bin_i - 1])
rplt.fill_between(
ratioPlusError,
ratioMinusError,
axes,
hatch='//',
# facecolor = 'Black',
facecolor='None',
edgecolor='Grey',
alpha=1.,
linewidth=0.,
# zorder = len(histograms_) + 1
zorder=0)
if CMS.tight_layout:
plt.tight_layout()
for save in save_as:
if save == 'root':
saveHistogramsToROOTFile(
data, stack,
save_folder + histogram_properties.name + '.' + save)
else:
plt.savefig(save_folder + histogram_properties.name + '.' + save)
plt.close()
def draw_data_mc(tf, hname, processes, signal_processes, **kwargs):
"""
Given a root file in the combine datacard format, draws a data/mc histogram,
ratio and systematic band.
tf (TFile): input root file
hname (string): name of histogram, example "sl_jge6_tge4/jetsByPt_0_pt"
processes (list of (str, str) tuples): processes to use for the plot. First
argument of tuple is the name in the file, second the name on the plot.
The order here defines the order of plotting (bottom to top). Assume that
signal is the first process.
signal_processes (list of str): process names to consider as signal
"""
# name (string) of the data process.
# Example: "data" (real data), "data_obs" (fake data)
#must exist in the input file
dataname = kwargs.get("dataname", "data_obs")
xlabel = kwargs.get("xlabel", escape_string(hname))
xunit = kwargs.get("xunit", "XUNIT")
ylabel = kwargs.get("ylabel", "auto")
rebin = kwargs.get("rebin", 1)
title_extended = kwargs.get("title_extended", "")
#legend properties
do_legend = kwargs.get("do_legend", True)
legend_loc = kwargs.get("legend_loc", (1.1,0.1))
legend_fontsize = kwargs.get("legend_fontsize", 6)
#Dictionary of sample (string) -> color (tuple of floats) to use as colors
#or "auto" to generate a sequence of colors
colors = kwargs.get("colors", "auto")
#True if you want to put the contents of the overflow bin into the last
#visible bin of the histogram, False otherwise
show_overflow = kwargs.get("show_overflow", False)
#function f: TH1D -> TH1D to apply on data to blind it.
blindFunc = kwargs.get("blindFunc", None)
#array of up-down pairs for systematic names to use for the systematic band,
#e.g.[("_CMS_scale_jUp", "_CMS_scale_jDown")]
systematics = kwargs.get("systematics", [])
histograms_nominal = getHistograms(tf, processes, hname)
if len(histograms_nominal) == 0:
raise KeyError("did not find any histograms for MC")
histograms_systematic = OrderedDict()
#get the systematically variated histograms
for systUp, systDown in systematics:
histograms_systematic[systUp] = getHistograms(tf, processes, hname+systUp)
histograms_systematic[systDown] = getHistograms(tf, processes, hname+systDown)
if len(histograms_systematic[systUp])==0 or len(histograms_systematic[systDown])==0:
print "Could not read histograms for {0}".format(hname+systUp)
processes_d = dict(processes)
for histo_dict in [histograms_nominal] + histograms_systematic.values():
for (proc, h) in histo_dict.items():
h.title = processes_d[proc] + " ({0:.1f})".format(h.Integral())
h.rebin(rebin)
if show_overflow:
fill_overflow(h)
c = plt.figure(figsize=(6,6))
#Create top panel
a1 = plt.axes([0.0, 0.22, 1.0, 0.8])
c.suptitle(r"$\textbf{CMS}$ preliminary $\sqrt{s} = 13$ TeV"+title_extended,
y=1.02, x=0.02,
horizontalalignment="left", verticalalignment="bottom", fontsize=16
)
stacked_hists = mc_stack(
histograms_nominal.values(),
histograms_systematic,
systematics,
colors = colors
)
#Create the normalized signal shape
histogram_signal = sum([histograms_nominal[sig] for sig in signal_processes])
histogram_total_mc = sum(histograms_nominal.values())
#hsig.Rebin(2)
if histogram_signal.Integral()>0:
histogram_signal.Scale(0.2 * histogram_total_mc.Integral() / histogram_signal.Integral())
histogram_signal.title = processes[0][1] + " norm"
histogram_signal.linewidth=2
histogram_signal.fillstyle = None
#draw the signal shape
hist([histogram_signal])
histogram_total_mc.title = "pseudodata"
histogram_total_mc.color = "black"
histogram_total_bkg = sum([
histograms_nominal[k] for k in histograms_nominal.keys()
if k not in signal_processes]
)
#Get the data histogram
data = None
if dataname != "":
data = tf.get(dataname + "/" + hname)
data.rebin(rebin)
if blindFunc:
data = blindFunc(data)
if show_overflow:
fill_overflow(data)
data.title = "data ({0:.2f})".format(data.Integral())
#set data error to 0 in case no data (FIXME)
for ibin in range(data.GetNbinsX()):
if data.GetBinContent(ibin) == 0:
data.SetBinError(ibin, 1)
errorbar(data)
if do_legend:
#create nice filled legend patches for all the processes
patches = []
if data:
dataline = mlines.Line2D([], [], color='black', marker='o', label=data.title)
patches += [dataline]
for (line1, line2), h in zip(stacked_hists["hists"], histograms_nominal.values()):
patch = mpatches.Patch(color=line1.get_color(), label=h.title)
patches += [patch]
patches += [mpatches.Patch(facecolor="none", edgecolor="black", label="stat", hatch="////////")]
patches += [mpatches.Patch(facecolor="none", edgecolor="gray", label="stat+syst", hatch=r"\\\\")]
plt.legend(handles=patches, loc=legend_loc, numpoints=1, prop={'size':legend_fontsize}, ncol=2, frameon=False)
#create an automatic bin width label on the y axis
if ylabel == "auto":
ylabel = "events / {0:.2f} {1}".format(histogram_signal.get_bin_width(1), xunit)
plt.ylabel(ylabel)
#hide x ticks on main panel
ticks = a1.get_xticks()
a1.get_xaxis().set_visible(False)
a1.set_ylim(bottom=0, top=1.1*a1.get_ylim()[1])
a1.grid(zorder=100000)
a2 = a1
#do ratio panel
if data:
a2 = plt.axes([0.0,0.0, 1.0, 0.18], sharex=a1)
plt.xlabel(xlabel)
a2.grid()
data_ratio = data.Clone()
data_ratio.Divide(histogram_total_mc)
#In case MC was empty, set data/mc ratio to 0
for ibin in range(data_ratio.GetNbinsX()):
bc = histogram_total_mc.GetBinContent(ibin)
if bc==0:
data_ratio.SetBinContent(ibin, 0)
#create also the variated band
bg_unc_u = stacked_hists["tot_u"]
bg_unc_d = stacked_hists["tot_d"]
bg_unc_u.Divide(stacked_hists["tot"])
bg_unc_d.Divide(stacked_hists["tot"])
bg_unc_usyst = stacked_hists["tot_usyst"]
bg_unc_dsyst = stacked_hists["tot_dsyst"]
bg_unc_usyst.Divide(stacked_hists["tot"])
bg_unc_dsyst.Divide(stacked_hists["tot"])
#blind the data also on the ratio
if blindFunc:
data_ratio = blindFunc(data_ratio)
errorbar(data_ratio)
fill_between(
bg_unc_u, bg_unc_d,
color="black", hatch="////////",
alpha=1.0, linewidth=0, facecolor="none", edgecolor="black", zorder=10,
)
fill_between(
bg_unc_usyst, bg_unc_dsyst,
color="gray", hatch=r"\\\\",
alpha=1.0, linewidth=0, facecolor="none", edgecolor="gray", zorder=10,
)
plt.title("data={0:.1f}\ MC={1:.1f}".format(
data.Integral(),
stacked_hists["tot"].Integral()
), x=0.01, y=0.8, fontsize=10, horizontalalignment="left"
)
plt.ylabel(r"$\frac{\mathrm{data}}{\mathrm{pred.}}$", fontsize=16)
plt.axhline(1.0, color="black")
a2.set_ylim(0, 2)
#hide last tick on ratio y axes
a2.set_yticks(a2.get_yticks()[:-1]);
a2.set_xticks(ticks);
return a1, a2, histograms_nominal, stacked_hists, histograms_systematic
def draw_data_mc(tf, hname, samples, **kwargs):
do_pseudodata = kwargs.get("do_pseudodata", False)
dataname = kwargs.get("dataname", None)
xlabel = kwargs.get("xlabel", hname.replace("_", " "))
xunit = kwargs.get("xunit", "XUNIT")
ylabel = kwargs.get("ylabel", "auto")
rebin = kwargs.get("rebin", 1)
title_extended = kwargs.get("title_extended", "")
do_legend = kwargs.get("do_legend", True)
legend_loc = kwargs.get("legend_loc", (1.1,0.1))
legend_fontsize = kwargs.get("legend_fontsize", 6)
colors = kwargs.get("colors", "auto")
show_overflow = kwargs.get("show_overflow", False)
blindFunc = kwargs.get("blindFunc", None)
#array of up-down pairs for systematic names, e.g. _CMS_scale_jUp/Down
systematics = kwargs.get("systematics", [])
hs = OrderedDict()
hs_syst = OrderedDict()
hs = getHistograms(tf, samples, hname)
#get the systematically variated histograms
for systUp, systDown in systematics:
hs_syst[systUp] = getHistograms(tf, samples, hname+systUp)
hs_syst[systDown] = getHistograms(tf, samples, hname+systDown)
if len(hs_syst[systUp])==0 or len(hs_syst[systDown])==0:
print "Could not read histograms for {0}".format(hname+systUp)
sample_d = dict(samples)
for hd in [hs] + hs_syst.values():
for (sample, h) in hd.items():
make_uoflow(h)
h.title = sample_d[sample] + " ({0:.1f})".format(h.Integral())
h.rebin(rebin)
if show_overflow:
fill_overflow(h)
c = plt.figure(figsize=(6,6))
if do_pseudodata or dataname:
a1 = plt.axes([0.0,0.22, 1.0, 0.8])
else:
a1 = plt.axes()
c.suptitle("$\\textbf{CMS}$ preliminary $\sqrt{s} = 13$ TeV"+title_extended,
y=1.02, x=0.02,
horizontalalignment="left", verticalalignment="bottom", fontsize=16
)
if len(hs) == 0:
raise KeyError("did not find any histograms for MC")
r = mc_stack(hs.values(), hs_syst, systematics, colors=colors)
#Create the normalized signal shape
hsig = hs[samples[0][0]].Clone()
tot_mc = sum(hs.values())
#hsig.Rebin(2)
if hsig.Integral()>0:
hsig.Scale(0.2 * tot_mc.Integral() / hsig.Integral())
hsig.title = samples[0][1] + " norm"
hsig.linewidth=2
hsig.fillstyle = None
hist([hsig])
tot_mc.title = "pseudodata"
tot_mc.color = "black"
tot_bg = sum([hs[k] for k in hs.keys() if "tth" not in k])
data = None
if do_pseudodata:
data = tot_mc.Clone()
data.title = "pseudodata"
if blindFunc:
data = blindFunc(data)
idata = data.Integral()
elif dataname:
datas = []
for dn in dataname:
try:
h = tf.get(dn + "/" + hname)
datas += [tf.get(dn + "/" + hname).Clone()]
except rootpy.io.file.DoesNotExist:
print "missing", dn, hname
if len(datas)>0:
data = sum(datas)
data.rebin(rebin)
else:
data = tot_mc.Clone()
data.Scale(0.0)
if blindFunc:
data = blindFunc(data)
data.title = "data ({0})".format(data.Integral())
idata = data.Integral()
if data and (blindFunc is None):
if show_overflow:
fill_overflow(data)
for ibin in range(data.GetNbinsX()):
if data.GetBinContent(ibin) == 0:
data.SetBinError(ibin, 1)
errorbar(data)
if do_legend:
patches = []
if data:
dataline = mlines.Line2D([], [], color='black', marker='o', label=data.title)
patches += [dataline]
for line, h in zip(r["hists"], hs.values()):
#import pdb
#pdb.set_trace()
patch = mpatches.Patch(color=line[0].get_color(), label=h.title)
patches += [patch]
patches += [mpatches.Patch(facecolor="none", edgecolor="black", label="stat", hatch="////////")]
patches += [mpatches.Patch(facecolor="none", edgecolor="gray", label="stat+syst", hatch="\\\\\\\\")]
plt.legend(handles=patches, loc=legend_loc, numpoints=1, prop={'size':legend_fontsize}, ncol=2, frameon=False)
if ylabel == "auto":
ylabel = "events / {0:.2f} {1}".format(hs.values()[0].get_bin_width(1), xunit)
plt.ylabel(ylabel)
if not data:
plt.xlabel(xlabel)
#hide x ticks on main panel
ticks = a1.get_xticks()
if data:
a1.get_xaxis().set_visible(False)
#print ticks
a1.set_ylim(bottom=0, top=1.1*a1.get_ylim()[1])
a1.grid(zorder=100000)
a2 = a1
#do ratio panel
if data:
a2 = plt.axes([0.0,0.0, 1.0, 0.18], sharex=a1)
plt.xlabel(xlabel)
a2.grid()
data.Divide(tot_mc)
for ibin in range(data.GetNbinsX()):
bc = tot_mc.GetBinContent(ibin)
if bc==0:
data.SetBinContent(ibin, 0)
bg_unc_u = r["tot_u"]
bg_unc_d = r["tot_d"]
bg_unc_u.Divide(r["tot"])
bg_unc_d.Divide(r["tot"])
bg_unc_usyst = r["tot_usyst"]
bg_unc_dsyst = r["tot_dsyst"]
bg_unc_usyst.Divide(r["tot"])
bg_unc_dsyst.Divide(r["tot"])
if blindFunc:
data = blindFunc(data)
errorbar(data)
fill_between(
bg_unc_u, bg_unc_d,
color="black", hatch="////////",
alpha=1.0, linewidth=0, facecolor="none", edgecolor="black", zorder=10,
)
fill_between(
bg_unc_usyst, bg_unc_dsyst,
color="gray", hatch="\\\\\\\\",
alpha=1.0, linewidth=0, facecolor="none", edgecolor="gray", zorder=10,
)
plt.title("data={0:.1f}\ MC={1:.1f}".format(idata, r["tot"].Integral()), x=0.01, y=0.8, fontsize=10, horizontalalignment="left")
plt.ylabel("$\\frac{\mathrm{data}}{\mathrm{pred.}}$", fontsize=16)
plt.axhline(1.0, color="black")
a2.set_ylim(0, 2)
#hide last tick on ratio y axes
a2.set_yticks(a2.get_yticks()[:-1]);
a2.set_xticks(ticks);
return a1, a2, hs, r, hs_syst
def make_data_mc_comparison_plot( histograms = [],
histogram_lables = [],
histogram_colors = [],
histogram_properties = Histogram_properties(),
data_index = 0,
save_folder = 'plots/',
save_as = ['pdf'],
normalise = False,
show_ratio = False,
show_stat_errors_on_mc = False,
draw_vertical_line = 0,
systematics_for_ratio = None
):
save_folder = check_save_folder(save_folder)
# make copies in order not to mess with existing histograms
histograms_ = deepcopy(histograms)
stack = HistStack()
add_mc = stack.Add
for index, histogram in enumerate( histograms_ ):
label = histogram_lables[index]
color = histogram_colors[index]
histogram.SetTitle( label )
if normalise:
histogram.Sumw2()
if not index == data_index:
histogram.fillstyle = 'solid'
histogram.fillcolor = color
histogram.legendstyle = 'F'
add_mc( histogram )
data = histograms_[data_index]
data.SetMarkerSize( CMS.data_marker_size )
if normalise:
n_events_data = data.Integral()
n_events_mc = stack.Integral()
data.Scale( 1 / n_events_data )
stack.Scale( 1 / n_events_mc )
# plot with matplotlib
plt.figure( figsize = CMS.figsize, dpi = CMS.dpi, facecolor = CMS.facecolor )
axes=None
if show_ratio:
ratio = data.Clone( 'ratio' )
ratio.Divide( sum( stack.GetHists() ) )
ratio.SetMarkerSize( 3 )
gs = gridspec.GridSpec( 2, 1, height_ratios = [5, 1] )
axes = plt.subplot( gs[0] )
else:
axes = plt.axes()
if histogram_properties.set_log_y:
axes.set_yscale( 'log', nonposy = "clip" )
axes.set_ylim( ymin = 1e-2 )
mc_error = histogram_properties.mc_error
if mc_error > 0:
stack_lower = sum( stack.GetHists() )
stack_upper = stack_lower.Clone( 'upper' )
stack_lower.Scale( 1 - mc_error )
stack_upper.Scale( 1 + mc_error )
rplt.fill_between( stack_upper,
stack_lower, axes, facecolor = '0.75',
alpha = 0.5, hatch = '/',
zorder = len(histograms_) + 1 )
if not mc_error > 0 and show_stat_errors_on_mc:
stack_lower = sum( stack.GetHists() )
mc_errors = list( stack_lower.yerravg() )
stack_upper = stack_lower.Clone( 'upper' )
for bin_i in range( 1, stack_lower.GetNbinsX() ):
stack_lower.SetBinContent( bin_i, stack_lower.GetBinContent( bin_i ) - mc_errors[bin_i - 1] )
stack_upper.SetBinContent( bin_i, stack_upper.GetBinContent( bin_i ) + mc_errors[bin_i - 1] )
rplt.fill_between( stack_upper, stack_lower, axes, facecolor = '0.75',
alpha = 0.5, hatch = '/',
zorder = len(histograms_) + 1 )
# a comment on zorder: the MC stack should be always at the very back (z = 1),
# then the MC error (z = len(histograms_) + 1) and finally the data
# (z = len(histograms_) + 2)
rplt.hist( stack, stacked = True, axes = axes, zorder = 1 )
rplt.errorbar( data, emptybins = histogram_properties.emptybins, axes = axes,
xerr = histogram_properties.xerr,
elinewidth = 2, capsize = 10, capthick = 2,
zorder = len(histograms_) + 2 )
# put legend into the correct order (data is always first!)
handles, labels = axes.get_legend_handles_labels()
data_label_index = labels.index( 'data' )
data_handle = handles[data_label_index]
labels.remove( 'data' )
handles.remove( data_handle )
labels.insert( 0, 'data' )
handles.insert( 0, data_handle )
if mc_error > 0 or ( not mc_error > 0 and show_stat_errors_on_mc ):
p1 = Rectangle( ( 0, 0 ), 1, 1, fc = "0.75", alpha = 0.5, hatch = '/' )
handles.append( p1 )
labels.append( histogram_properties.mc_errors_label )
l1 = axes.legend( handles, labels, numpoints = 1,
frameon = histogram_properties.legend_color,
bbox_to_anchor = histogram_properties.legend_location,
bbox_transform=plt.gcf().transFigure,
prop = CMS.legend_properties,
ncol = histogram_properties.legend_columns )
l1.set_zorder(102)
set_labels( plt, histogram_properties, show_x_label = not show_ratio, axes = axes )
x_limits = histogram_properties.x_limits
y_limits = histogram_properties.y_limits
if len( x_limits ) >= 2:
axes.set_xlim( xmin = x_limits[0], xmax = x_limits[-1] )
if len( y_limits ) >= 2:
axes.set_ylim( ymin = y_limits[0], ymax = y_limits[-1] )
else:
y_max = get_best_max_y(histograms_, x_limits=x_limits) * histogram_properties.y_max_scale
axes.set_ylim( ymin = 0, ymax = y_max )
if histogram_properties.set_log_y:
if not len( y_limits ) == 2: # if not user set y-limits, set default
axes.set_ylim( ymin = 1e-1 )
#draw a red vertical line if needed:
if draw_vertical_line != 0:
plt.axvline(x = draw_vertical_line, color = 'red', linewidth = 3)
if show_ratio:
plt.setp( axes.get_xticklabels(), visible = False )
ax1 = plt.subplot( gs[1] )
ax1.minorticks_on()
ax1.grid( True, 'major', linewidth = 1 )
ax1.axhline(y=1, linewidth = 1)
set_labels( plt, histogram_properties, show_x_label = True, show_title = False )
plt.ylabel( r'$\frac{\mathrm{data}}{\mathrm{pred.}}$', CMS.y_axis_title )
ax1.yaxis.set_label_coords(-0.115, 0.8)
rplt.errorbar( ratio, emptybins = histogram_properties.emptybins, axes = ax1,
xerr = histogram_properties.xerr,
elinewidth = 1.5, capsize = 5, capthick = 1.5 )
if len( x_limits ) >= 2:
ax1.set_xlim( xmin = x_limits[0], xmax = x_limits[-1] )
if len( histogram_properties.ratio_y_limits ) >= 2:
ax1.set_ylim( ymin = histogram_properties.ratio_y_limits[0],
ymax = histogram_properties.ratio_y_limits[-1] )
# dynamic tick placement
adjust_ratio_ticks(ax1.yaxis, n_ticks = 3, y_limits = histogram_properties.ratio_y_limits)
if histogram_properties.integerXVariable :
ax1.tick_params(axis='x',which='minor',bottom='off',top='off')
if systematics_for_ratio != None:
plusErrors = [x+1 for x in systematics_for_ratio]
minusErrors = [1-x for x in systematics_for_ratio]
print plusErrors
print minusErrors
ratioPlusError = ratio.Clone( 'plus' )
ratioMinusError = ratio.Clone( 'minus' )
for bin_i in range( 1, ratioPlusError.GetNbinsX()+1 ):
ratioPlusError.SetBinContent( bin_i, plusErrors[bin_i-1] )
ratioMinusError.SetBinContent( bin_i, minusErrors[bin_i-1] )
rplt.fill_between( ratioPlusError, ratioMinusError, axes,
alpha = 0.3, hatch = '//',
facecolor = 'Black',
zorder = len(histograms_) + 1 )
if CMS.tight_layout:
plt.tight_layout()
for save in save_as:
if save == 'root':
saveHistogramsToROOTFile( data, stack,save_folder + histogram_properties.name + '.' + save )
else:
plt.savefig( save_folder + histogram_properties.name + '.' + save )
plt.close()
def make_plots( histograms, category, output_folder, histname, show_ratio = False, show_generator_ratio = False, show_before_unfolding = False, utype = 'normalised', preliminary=True ):
global variable, phase_space
channel = ''
if 'electron' in histname:
channel = 'electron'
elif 'muon' in histname:
channel = 'muon'
else:
channel = 'combined'
# Initailise data histograms
hist_data = histograms['unfolded']
hist_data.markersize = 2
hist_data.marker = 'o'
if category == 'central':
hist_data_with_systematics = histograms['unfolded_with_systematics']
hist_data_with_systematics.markersize = 2
hist_data_with_systematics.marker = 'o'
# Create base figure to be plotted
plt.figure( figsize = CMS.figsize, dpi = CMS.dpi, facecolor = CMS.facecolor )
# Split into 3 for MC/Data ratio and generator ratio and plot
if show_ratio and show_generator_ratio:
gs = gridspec.GridSpec( 3, 1, height_ratios = [5, 1, 1] )
axes = plt.subplot( gs[0] )
# Split into 2 for MC/Data ratio or generator Ratio and plot
elif show_ratio or show_generator_ratio:
gs = gridspec.GridSpec( 2, 1, height_ratios = [4, 1] )
axes = plt.subplot( gs[0] )
# Just 1 for plot and setup x axis labels
else:
axes = plt.axes()
x_label = '${}$'.format(variables_latex[variable])
if variable in ['HT', 'ST', 'MET', 'WPT', 'lepton_pt']:
x_label += ' [GeV]'
plt.xlabel( x_label, CMS.x_axis_title )
# set y axis x-section labels
y_label = ''
xsectionUnit = ''
if utype == 'absolute':
y_label = r'$\frac{d\sigma}{d' + variables_latex[variable] + '}\ '
xsectionUnit = 'pb'
else :
y_label = r'$\frac{1}{\sigma} \frac{d\sigma}{d' + variables_latex[variable] + '}\ '
if variable in ['HT', 'ST', 'MET', 'WPT', 'lepton_pt']:
if xsectionUnit is '':
y_label += '\scriptstyle(\mathrm{GeV}^{-1})$'
pass
else:
y_label += '\scriptstyle(\mathrm{'
y_label += xsectionUnit
y_label += '}\ \mathrm{GeV}^{-1})$'
elif xsectionUnit is not '':
y_label += '\scriptstyle(\mathrm{'
y_label += xsectionUnit
y_label += '})$'
plt.ylabel( y_label, CMS.y_axis_title )
# Set up ticks on axis. Minor ticks on axis for non NJet variables
plt.tick_params( **CMS.axis_label_major )
if not variable in ['NJets']:
axes.minorticks_on()
plt.tick_params( **CMS.axis_label_minor )
# Set raw unfolded data with stat+unfolding uncertianty to be visible
hist_data.visible = True
# axes.set_yscale('log')
# Set raw unfolded data with systematic uncertianty to be visible
# label = 'do_not_show' = do not show in legend
if category == 'central':
hist_data_with_systematics.visible = True
rplt.errorbar(
hist_data_with_systematics,
axes = axes,
label = 'do_not_show',
xerr = None,
capsize = 0,
elinewidth = 2,
zorder = len( histograms ) + 1
)
# Show stat+unf uncertainty on plot
rplt.errorbar( hist_data,
axes = axes,
label = 'do_not_show',
xerr = None,
capsize = 15,
capthick = 3,
elinewidth = 2,
zorder = len( histograms ) + 2
)
# And one for a nice legend entry
rplt.errorbar( hist_data,
axes = axes,
label = 'data',
xerr = None,
yerr = False,
capsize = 0,
elinewidth = 2,
zorder = len( histograms ) + 3
)
dashes = {}
for key, hist in sorted( histograms.items() ):
zorder = sorted( histograms, reverse = False ).index( key )
# Ordering such that systematic uncertainties are plotted first then central powhegPythia then data
if key == 'TTJets_powhegPythia8' and zorder != len(histograms) - 3:
zorder = len(histograms) - 3
elif key != 'TTJets_powhegPythia8' and not 'unfolded' in key:
while zorder >= len(histograms) - 3:
zorder = zorder - 1
# Colour and style of MC hists
if not 'unfolded' in key and not 'measured' in key:
hist.linewidth = 4
linestyle = None
if 'powhegPythia8' in key:
linestyle = 'solid'
dashes[key] = None
hist.SetLineColor( 633 )
elif 'powhegHerwig' in key or 'isr' in key or 'hdamp' in key:
hist.SetLineColor( kBlue )
dashes[key] = [25,5,5,5,5,5,5,5]
elif 'amcatnloPythia8' in key or 'fsr' in key or 'scale' in key:
hist.SetLineColor( 807 )
dashes[key] = [20,5]
elif 'madgraphMLM' in key or 'renormalisation' in key or 'topPt' in key:
hist.SetLineColor( 417 )
dashes[key] = [5,5]
elif 'factorisation' in key or 'ue' in key:
hist.SetLineColor( 100 )
dashes[key] = [10,10]
elif 'combined' in key or 'semiLepBr' in key:
hist.SetLineColor( 200 )
dashes[key] = [20,10]
elif 'semiLepBr' in key or 'mass' in key:
hist.SetLineColor( 300 )
dashes[key] = [20,10,10,10]
elif 'frag' in key or 'Frag' in key:
hist.SetLineColor( 400 )
dashes[key] = [10,10,5,5]
elif 'erdOn' in key:
hist.SetLineColor( 500 )
dashes[key] = [10,10,5,5,5,5]
if linestyle != None:
hist.linestyle = linestyle
# Add hist to plot
line, h = rplt.hist( hist, axes = axes, label = measurements_latex[key], zorder = zorder )
# Set the dashes and lines
if dashes[key] != None:
line.set_dashes(dashes[key])
h.set_dashes(dashes[key])
handles, labels = axes.get_legend_handles_labels()
# Making data first in the legend
data_label_index = labels.index( 'data' )
data_handle = handles[data_label_index]
labels.remove( 'data' )
handles.remove( data_handle )
labels.insert( 0, 'data' )
handles.insert( 0, data_handle )
# Order the rest of the labels in the legend
new_handles, new_labels = [], []
zipped = dict( zip( labels, handles ) )
labelOrder = ['data',
measurements_latex['TTJets_powhegPythia8'],
measurements_latex['TTJets_amcatnloPythia8'],
measurements_latex['TTJets_powhegHerwig'],
measurements_latex['TTJets_madgraphMLM'],
measurements_latex['TTJets_scaleup'],
measurements_latex['TTJets_scaledown'],
measurements_latex['TTJets_massup'],
measurements_latex['TTJets_massdown'],
measurements_latex['TTJets_ueup'],
measurements_latex['TTJets_uedown'],
measurements_latex['TTJets_fsrup'],
measurements_latex['TTJets_fsrdown'],
measurements_latex['TTJets_isrdown'],
measurements_latex['TTJets_isrup'],
# measurements_latex['TTJets_alphaSup'],
# measurements_latex['TTJets_alphaSdown'],
measurements_latex['TTJets_topPt'],
measurements_latex['TTJets_factorisationup'],
measurements_latex['TTJets_factorisationdown'],
measurements_latex['TTJets_renormalisationup'],
measurements_latex['TTJets_renormalisationdown'],
measurements_latex['TTJets_combinedup'],
measurements_latex['TTJets_combineddown'],
measurements_latex['TTJets_hdampup'],
measurements_latex['TTJets_hdampdown'],
measurements_latex['TTJets_erdOn'],
measurements_latex['TTJets_QCDbased_erdOn'],
measurements_latex['TTJets_GluonMove'],
measurements_latex['TTJets_semiLepBrup'],
measurements_latex['TTJets_semiLepBrdown'],
measurements_latex['TTJets_fragup'],
measurements_latex['TTJets_fragdown'],
measurements_latex['TTJets_petersonFrag'],
]
for label in labelOrder:
if label in labels:
new_handles.append(zipped[label])
if label == 'data':
new_labels.append(measurements_latex['data'])
else:
new_labels.append(label)
# Location of the legend
legend_location = (0.95, 0.82)
prop = CMS.legend_properties
if variable == 'MT':
legend_location = (0.05, 0.82)
elif variable == 'ST':
legend_location = (0.97, 0.82)
elif variable == 'WPT':
legend_location = (1.0, 0.84)
elif variable == 'abs_lepton_eta' or variable =='abs_lepton_eta_coarse':
legend_location = (0.97, 0.87)
elif variable == 'NJets':
# Reduce size of NJets legend
prop = {'size':30}
# Add legend to plot
plt.legend( new_handles, new_labels,
numpoints = 1,
prop = prop,
frameon = False,
bbox_to_anchor=legend_location,
bbox_transform=plt.gcf().transFigure
)
# Title and CMS labels
# note: fontweight/weight does not change anything as we use Latex text!!!
label, channel_label = get_cms_labels( channel )
plt.title( label,loc='right', **CMS.title )
# Locations of labels
logo_location = (0.05, 0.97)
channel_location = ( 0.05, 0.9)
if preliminary:
prelim_location = (0.05, 0.9)
channel_location = ( 0.5, 0.97)
# preliminary
plt.text(prelim_location[0], prelim_location[1],
r"\emph{Preliminary}",
transform=axes.transAxes,
fontsize=42,
verticalalignment='top',
horizontalalignment='left'
)
# if variable == 'WPT':
# logo_location = (0.05, 0.97)
# prelim_location = (0.05, 0.9)
# channel_location = (0.5, 0.97)
# elif variable == 'abs_lepton_eta':
# logo_location = (0.05, 0.97)
# prelim_location = (0.05, 0.9)
# channel_location = ( 0.5, 0.97)
# Add labels to plot
plt.text(logo_location[0], logo_location[1],
r"\textbf{CMS}",
transform=axes.transAxes,
fontsize=42,
verticalalignment='top',
horizontalalignment='left'
)
# channel text
plt.text(channel_location[0], channel_location[1],
r"%s"%channel_label,
transform=axes.transAxes,
fontsize=40,
verticalalignment='top',
horizontalalignment='left'
)
# Set y limits on plot
ylim = axes.get_ylim()
if ylim[0] < 0:
axes.set_ylim( ymin = 0.)
axes.set_ylim(ymax = ylim[1]*1.1)
if variable == 'abs_lepton_eta' or variable == 'abs_lepton_eta_coarse':
axes.set_ylim(ymax = ylim[1]*1.6)
# Now to show either of the ratio plots
if show_ratio:
# Set previous x axis ticks and labels to invisible
plt.setp( axes.get_xticklabels(), visible = False )
# Go to ratio subplot
ax1 = plt.subplot( gs[1] )
# setting the x_limits identical to the main plot
x_limits = axes.get_xlim()
ax1.set_xlim(x_limits)
# Setting tick marks
ax1.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
plt.tick_params( **CMS.axis_label_major )
# if not variable in ['NJets']:
# ax1.minorticks_on()
# ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
# plt.tick_params( **CMS.axis_label_minor )
# x axis labels as before
x_label = '${}$'.format(variables_latex[variable])
if variable in ['HT', 'ST', 'MET', 'WPT', 'lepton_pt']:
x_label += ' (GeV)'
if not show_generator_ratio:
plt.xlabel( x_label, CMS.x_axis_title )
y_label = '$\\displaystyle\\frac{\\mathrm{pred.}}{\\mathrm{data}}$'
plt.ylabel( y_label, CMS.y_axis_title_small )
ax1.yaxis.set_label_coords(-0.115, 0.7)
# Draw a horizontal line at y=1 for data
plt.axhline(y = 1, color = 'black', linewidth = 2)
# Create ratios and plot to subplot
for key, hist in sorted( histograms.iteritems() ):
if not 'unfolded' in key and not 'measured' in key:
ratio = hist.Clone()
ratio.Divide( hist_data )
line, h = rplt.hist( ratio, axes = ax1, label = 'do_not_show' )
if dashes[key] != None:
line.set_dashes(dashes[key])
h.set_dashes(dashes[key])
# Now for the error bands
stat_lower = hist_data.Clone()
stat_upper = hist_data.Clone()
syst_lower = hist_data.Clone()
syst_upper = hist_data.Clone()
# Plot relative error bands on data in the ratio plot
stat_errors = graph_to_value_errors_tuplelist(hist_data)
if category == 'central':
syst_errors = graph_to_value_errors_tuplelist(hist_data_with_systematics)
for bin_i in range( 1, hist_data.GetNbinsX() + 1 ):
stat_value, stat_error, _ = stat_errors[bin_i-1]
stat_rel_error = stat_error/stat_value
stat_lower.SetBinContent( bin_i, 1 - stat_rel_error )
stat_upper.SetBinContent( bin_i, 1 + stat_rel_error )
if category == 'central':
syst_value, syst_error_down, syst_error_up = syst_errors[bin_i-1]
syst_rel_error_down = syst_error_down/syst_value
syst_rel_error_up = syst_error_up/syst_value
syst_lower.SetBinContent( bin_i, 1 - syst_rel_error_down )
syst_upper.SetBinContent( bin_i, 1 + syst_rel_error_up )
# Colour
if category == 'central':
rplt.fill_between(
syst_lower,
syst_upper,
ax1,
color = 'gold'
)
rplt.fill_between(
stat_upper,
stat_lower,
ax1,
color = '0.75',
)
# Add legend
loc = 'upper left'
if variable == 'MET':
loc = 'lower left'
elif variable == 'HT':
loc = 'lower center'
elif variable == 'ST':
loc = 'lower center'
elif variable == 'WPT':
loc = 'lower left'
elif variable == 'NJets':
loc = 'lower left'
elif variable == 'abs_lepton_eta' or variable == 'abs_lepton_eta_coarse':
loc = 'upper left'
elif variable == 'lepton_pt':
loc = 'lower left'
# legend for ratio plot
p_stat = mpatches.Patch(facecolor='0.75', label='Stat.', edgecolor='black' )
p_stat_and_syst = mpatches.Patch(facecolor='gold', label=r'Stat. $\oplus$ Syst.', edgecolor='black' )
l1 = ax1.legend(
handles = [p_stat, p_stat_and_syst],
loc = loc,
frameon = False,
prop = {'size':26},
ncol = 2
)
ax1.add_artist(l1)
# Setting y limits and tick parameters
if variable == 'MET':
ax1.set_ylim( ymin = 0.6, ymax = 1.4 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
if variable == 'MT':
ax1.set_ylim( ymin = 0.8, ymax = 1.2 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'HT':
ax1.set_ylim( ymin = 0.6, ymax = 1.4 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'ST':
ax1.set_ylim( ymin = 0.6, ymax = 1.4 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'WPT':
ax1.set_ylim( ymin = 0.6, ymax = 1.4 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'NJets':
ax1.set_ylim( ymin = 0.6, ymax = 1.4 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'abs_lepton_eta' or variable == 'abs_lepton_eta_coarse':
ax1.set_ylim( ymin = 0.6, ymax = 1.4 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'lepton_pt':
ax1.set_ylim( ymin = 0.6, ymax = 1.4 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
if show_generator_ratio:
ax2 = None
# Remove Data/MC Ratio Axis
if show_ratio:
plt.setp( ax1.get_xticklabels(), visible = False )
ax2 = plt.subplot( gs[2] )
else:
plt.setp( axes.get_xticklabels(), visible = False )
ax2 = plt.subplot( gs[1] )
# setting the x_limits identical to the main plot
x_limits = axes.get_xlim()
ax2.set_xlim(x_limits)
# Setting ticks
ax2.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
plt.tick_params( **CMS.axis_label_major )
if not variable in ['NJets']:
ax2.minorticks_on()
ax2.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
plt.tick_params( **CMS.axis_label_minor )
# x axis labels as before
x_label = '${}$'.format(variables_latex[variable])
if variable in ['HT', 'ST', 'MET', 'WPT', 'lepton_pt']:
x_label += ' [GeV]'
plt.xlabel( x_label, CMS.x_axis_title )
y_label = 'Ratio to \n$' + measurements_latex['TTJets_powhegPythia8'] + '$'
plt.ylabel( y_label, CMS.y_axis_title_tiny )
#draw a horizontal line at y=1 for central MC
plt.axhline(y = 1, color = 'black', linewidth = 2)
central_mc = histograms['TTJets_powhegPythia8']
for key, hist in sorted( histograms.iteritems() ):
if not 'measured' in key:
ratio = None
if not 'unfolded_with_systematics' in key:
ratio = hist.Clone()
ratio.Divide( central_mc ) #divide by central mc sample
else:
ratio = central_mc.Clone()
syst_errors = graph_to_value_errors_tuplelist(hist)
for bin_i in range( 1, ratio.GetNbinsX() + 1 ):
syst_value, syst_error_down, syst_error_up = syst_errors[bin_i-1]
mc = central_mc.GetBinContent(bin_i)
data = list(hist.y())[bin_i-1]
ratio.SetBinContent( bin_i, data / mc)
ratio.SetBinError( bin_i, syst_error_down / mc )
if not 'unfolded' in key:
line, h = rplt.hist( ratio, axes = ax2, label = 'do_not_show' )
if dashes[key] != None:
line.set_dashes(dashes[key])
h.set_dashes(dashes[key])
elif 'unfolded_with_systematics' in key:
ratio.markersize = 2
ratio.marker = 'o'
ratio.color = 'black'
rplt.errorbar(
ratio,
axes = ax2,
label = 'do_not_show',
xerr = None,
capsize = 0,
elinewidth = 2,
zorder = len( histograms ) + 10
)
else:
ratio.markersize = 2
ratio.marker = 'o'
rplt.errorbar( ratio,
axes = ax2,
label = 'do_not_show',
xerr = None,
capsize = 15,
capthick = 3,
elinewidth = 2,
zorder = len( histograms ) + 9
)
if variable == 'MET':
ax2.set_ylim( ymin = 0.8, ymax = 1.2 )
ax2.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
ax2.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
if variable == 'MT':
ax2.set_ylim( ymin = 0.8, ymax = 1.2 )
ax2.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax2.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'HT':
ax2.set_ylim( ymin = 0.7, ymax = 1.3 )
ax2.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax2.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'ST':
ax2.set_ylim( ymin = 0.7, ymax = 1.5 )
ax2.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
ax2.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'WPT':
ax2.set_ylim( ymin = 0.8, ymax = 1.2 )
ax2.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
ax2.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'NJets':
ax2.set_ylim( ymin = 0.7, ymax = 1.5 )
elif variable == 'abs_lepton_eta' or variable == 'abs_lepton_eta_coarse':
ax2.set_ylim( ymin = 0.8, ymax = 1.2 )
ax2.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax2.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'lepton_pt':
ax2.set_ylim( ymin = 0.8, ymax = 1.3 )
ax2.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax2.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
if CMS.tight_layout:
plt.tight_layout()
# Save the plots
path = '{output_folder}/xsections/{phaseSpace}/{variable}/'
path = path.format(
output_folder = output_folder,
phaseSpace = phase_space,
variable = variable
)
make_folder_if_not_exists( path )
for output_format in output_formats:
filename = path + '/' + histname + '.' + output_format
plt.savefig( filename )
del hist_data
if 'central' in category: del hist_data_with_systematics
plt.close()
gc.collect()
return
def make_data_mc_comparison_plot( histograms = [],
histogram_lables = [],
histogram_colors = [],
histogram_properties = Histogram_properties(),
data_index = 0,
save_folder = 'plots/',
save_as = ['pdf', 'png'],
normalise = False,
show_ratio = False,
show_stat_errors_on_mc = False,
draw_vertical_line = 0,
):
# make copies in order not to mess with existing histograms
histograms_ = deepcopy(histograms)
stack = HistStack()
add_mc = stack.Add
for index, histogram in enumerate( histograms_ ):
label = histogram_lables[index]
color = histogram_colors[index]
histogram.SetTitle( label )
if normalise:
histogram.Sumw2()
if not index == data_index:
histogram.fillstyle = 'solid'
histogram.fillcolor = color
histogram.legendstyle = 'F'
add_mc( histogram )
data = histograms_[data_index]
data.SetMarkerSize( CMS.data_marker_size )
if normalise:
n_events_data = data.Integral()
n_events_mc = stack.Integral()
data.Scale( 1 / n_events_data )
stack.Scale( 1 / n_events_mc )
# plot with matplotlib
plt.figure( figsize = CMS.figsize, dpi = CMS.dpi, facecolor = CMS.facecolor )
if show_ratio:
ratio = data.Clone( 'ratio' )
ratio.Divide( sum( stack.GetHists() ) )
ratio.SetMarkerSize( 3 )
gs = gridspec.GridSpec( 2, 1, height_ratios = [5, 1] )
axes = plt.subplot( gs[0] )
else:
axes = plt.axes()
if histogram_properties.set_log_y:
axes.set_yscale( 'log', nonposy = "clip" )
axes.set_ylim( ymin = 1e-2 )
mc_error = histogram_properties.mc_error
if mc_error > 0:
stack_lower = sum( stack.GetHists() )
stack_upper = stack_lower.Clone( 'upper' )
stack_lower.Scale( 1 - mc_error )
stack_upper.Scale( 1 + mc_error )
rplt.fill_between( stack_upper,
stack_lower, axes, facecolor = '0.75',
alpha = 0.5, hatch = '/',
zorder = len(histograms_) + 1 )
if not mc_error > 0 and show_stat_errors_on_mc:
stack_lower = sum( stack.GetHists() )
mc_errors = list( stack_lower.yerravg() )
stack_upper = stack_lower.Clone( 'upper' )
for bin_i in range( 1, stack_lower.GetNbinsX() ):
stack_lower.SetBinContent( bin_i, stack_lower.GetBinContent( bin_i ) - mc_errors[bin_i - 1] )
stack_upper.SetBinContent( bin_i, stack_upper.GetBinContent( bin_i ) + mc_errors[bin_i - 1] )
rplt.fill_between( stack_upper, stack_lower, axes, facecolor = '0.75',
alpha = 0.5, hatch = '/',
zorder = len(histograms_) + 1 )
# a comment on zorder: the MC stack should be always at the very back (z = 1),
# then the MC error (z = len(histograms_) + 1) and finally the data
# (z = len(histograms_) + 2)
rplt.hist( stack, stacked = True, axes = axes, zorder = 1 )
rplt.errorbar( data, xerr = False, emptybins = False, axes = axes,
elinewidth = 2, capsize = 10, capthick = 2,
zorder = len(histograms_) + 2 )
# put legend into the correct order (data is always first!)
handles, labels = axes.get_legend_handles_labels()
data_label_index = labels.index( 'data' )
data_handle = handles[data_label_index]
labels.remove( 'data' )
handles.remove( data_handle )
labels.insert( 0, 'data' )
handles.insert( 0, data_handle )
if mc_error > 0 or ( not mc_error > 0 and show_stat_errors_on_mc ):
p1 = Rectangle( ( 0, 0 ), 1, 1, fc = "0.75", alpha = 0.5, hatch = '/' )
handles.append( p1 )
labels.append( histogram_properties.mc_errors_label )
plt.legend( handles, labels, numpoints = 1, loc = histogram_properties.legend_location,
prop = CMS.legend_properties, ncol = histogram_properties.legend_columns ).set_zorder(102)
set_labels( plt, histogram_properties, show_x_label = not show_ratio )
x_limits = histogram_properties.x_limits
y_limits = histogram_properties.y_limits
if len( x_limits ) == 2:
axes.set_xlim( xmin = x_limits[0], xmax = x_limits[1] )
if len( y_limits ) == 2:
axes.set_ylim( ymin = y_limits[0], ymax = y_limits[1] )
else:
axes.set_ylim( ymin = 0 )
if histogram_properties.set_log_y:
if not len( y_limits ) == 2: # if not user set y-limits, set default
axes.set_ylim( ymin = 1e-1 )
#draw a red vertical line if needed:
if draw_vertical_line != 0:
plt.axvline(x = draw_vertical_line, color = 'red', linewidth = 3)
if show_ratio:
plt.setp( axes.get_xticklabels(), visible = False )
ax1 = plt.subplot( gs[1] )
ax1.minorticks_on()
ax1.grid( True, 'major', linewidth = 1 )
ax1.yaxis.set_major_locator( MultipleLocator( 1.0 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.5 ) )
set_labels( plt, histogram_properties, show_x_label = True, show_title = False )
plt.ylabel( 'data/MC', CMS.y_axis_title )
rplt.errorbar( ratio, xerr = True, emptybins = False, axes = ax1 )
if len( x_limits ) == 2:
ax1.set_xlim( xmin = x_limits[0], xmax = x_limits[1] )
ax1.set_ylim( ymin = 0, ymax = 2 )
if CMS.tight_layout:
plt.tight_layout()
for save in save_as:
plt.savefig( save_folder + histogram_properties.name + '.' + save )
plt.close()
def make_plots( histograms, category, output_folder, histname, show_ratio = True, show_before_unfolding = False ):
global variable, k_values
channel = 'electron'
if 'electron' in histname:
channel = 'electron'
elif 'muon' in histname:
channel = 'muon'
else:
channel = 'combined'
# plot with matplotlib
hist_data = histograms['unfolded']
if category == 'central':
hist_data_with_systematics = histograms['unfolded_with_systematics']
hist_measured = histograms['measured']
hist_data.markersize = 2
hist_data.marker = 'o'
if category == 'central':
hist_data_with_systematics.markersize = 2
hist_data_with_systematics.marker = 'o'
hist_measured.markersize = 2
hist_measured.marker = 'o'
hist_measured.color = 'red'
plt.figure( figsize = CMS.figsize, dpi = CMS.dpi, facecolor = CMS.facecolor )
if show_ratio:
gs = gridspec.GridSpec( 2, 1, height_ratios = [5, 1] )
axes = plt.subplot( gs[0] )
else:
axes = plt.axes()
plt.xlabel( '$%s$ [GeV]' % variables_latex[variable], CMS.x_axis_title )
axes.minorticks_on()
plt.ylabel( r'$\frac{1}{\sigma} \frac{d\sigma}{d' + variables_latex[variable] + '} \left[\mathrm{GeV}^{-1}\\right]$', CMS.y_axis_title )
plt.tick_params( **CMS.axis_label_major )
plt.tick_params( **CMS.axis_label_minor )
hist_data.visible = True
if category == 'central':
hist_data_with_systematics.visible = True
rplt.errorbar( hist_data_with_systematics, axes = axes, label = 'do_not_show', xerr = None, capsize = 0, elinewidth = 2, zorder = len( histograms ) + 1 )
rplt.errorbar( hist_data, axes = axes, label = 'do_not_show', xerr = None, capsize = 15, capthick = 3, elinewidth = 2, zorder = len( histograms ) + 2 )
rplt.errorbar( hist_data, axes = axes, label = 'data', xerr = None, yerr = False, zorder = len( histograms ) + 3 ) # this makes a nicer legend entry
if show_before_unfolding:
rplt.errorbar( hist_measured, axes = axes, label = 'data (before unfolding)', xerr = None, zorder = len( histograms ) )
for key, hist in sorted( histograms.iteritems() ):
if not 'unfolded' in key and not 'measured' in key:
hist.linewidth = 4
# setting colours
if 'POWHEG_PYTHIA' in key or 'matchingdown' in key:
hist.linestyle = 'longdashdot'
hist.SetLineColor( kBlue )
elif 'POWHEG_HERWIG' in key or 'scaledown' in key:
hist.linestyle = 'dashed'
hist.SetLineColor( kGreen )
elif 'MADGRAPH_ptreweight' in key:
hist.linestyle = 'dashed'
hist.SetLineColor( kBlack )
elif 'MADGRAPH' in key:
hist.linestyle = 'solid'
hist.SetLineColor( kRed + 1 )
elif 'matchingup' in key:
hist.linestyle = 'verylongdashdot'
hist.linecolor = 'orange'
elif 'MCATNLO' in key or 'scaleup' in key:
hist.linestyle = 'dotted'
hist.SetLineColor( kMagenta + 3 )
rplt.hist( hist, axes = axes, label = measurements_latex[key], zorder = sorted( histograms, reverse = True ).index( key ) )
handles, labels = axes.get_legend_handles_labels()
# making data first in the list
data_label_index = labels.index( 'data' )
data_handle = handles[data_label_index]
labels.remove( 'data' )
handles.remove( data_handle )
labels.insert( 0, 'unfolded data' )
handles.insert( 0, data_handle )
new_handles, new_labels = [], []
for handle, label in zip( handles, labels ):
if not label == 'do_not_show':
new_handles.append( handle )
new_labels.append( label )
legend_location = (0.98, 0.88)
if variable == 'MT':
legend_location = (0.05, 0.88)
elif variable == 'ST':
legend_location = (0.95, 0.88)
plt.legend( new_handles, new_labels, numpoints = 1, prop = CMS.legend_properties, frameon = False, bbox_to_anchor=legend_location,
bbox_transform=plt.gcf().transFigure )
label, channel_label = get_cms_labels( channel )
# title
plt.title( label,loc='right', **CMS.title )
# CMS text
# note: fontweight/weight does not change anything as we use Latex text!!!
plt.text(0.05, 0.98, r"\textbf{CMS}", transform=axes.transAxes, fontsize=42,
verticalalignment='top',horizontalalignment='left')
# channel text
axes.text(0.95, 0.98, r"\emph{%s}" %channel_label, transform=axes.transAxes, fontsize=40,
verticalalignment='top',horizontalalignment='right')
if show_ratio:
plt.setp( axes.get_xticklabels(), visible = False )
ax1 = plt.subplot( gs[1] )
ax1.minorticks_on()
#ax1.grid( True, 'major', linewidth = 1 )
# setting the x_limits identical to the main plot
x_limits = axes.get_xlim()
ax1.set_xlim(x_limits)
ax1.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
plt.xlabel( '$%s$ [GeV]' % variables_latex[variable], CMS.x_axis_title )
plt.tick_params( **CMS.axis_label_major )
plt.tick_params( **CMS.axis_label_minor )
plt.ylabel( '$\\frac{\\textrm{pred.}}{\\textrm{data}}$', CMS.y_axis_title_small )
ax1.yaxis.set_label_coords(-0.115, 0.8)
#draw a horizontal line at y=1 for data
plt.axhline(y = 1, color = 'black', linewidth = 2)
for key, hist in sorted( histograms.iteritems() ):
if not 'unfolded' in key and not 'measured' in key:
ratio = hist.Clone()
ratio.Divide( hist_data ) #divide by data
rplt.hist( ratio, axes = ax1, label = 'do_not_show' )
stat_lower = hist_data.Clone()
stat_upper = hist_data.Clone()
syst_lower = hist_data.Clone()
syst_upper = hist_data.Clone()
# plot error bands on data in the ratio plot
for bin_i in range( 1, hist_data.GetNbinsX() + 1 ):
stat_errors = graph_to_value_errors_tuplelist(hist_data)
stat_lower.SetBinContent( bin_i, 1 - stat_errors[bin_i-1][1]/stat_errors[bin_i-1][0] )
stat_upper.SetBinContent( bin_i, 1 + stat_errors[bin_i-1][2]/stat_errors[bin_i-1][0] )
if category == 'central':
syst_errors = graph_to_value_errors_tuplelist(hist_data_with_systematics)
syst_lower.SetBinContent( bin_i, 1 - syst_errors[bin_i-1][1]/syst_errors[bin_i-1][0] )
syst_upper.SetBinContent( bin_i, 1 + syst_errors[bin_i-1][2]/syst_errors[bin_i-1][0] )
if category == 'central':
rplt.fill_between( syst_lower, syst_upper, ax1,
color = 'yellow', alpha = 0.5 )
rplt.fill_between( stat_upper, stat_lower, ax1, color = '0.75',
alpha = 0.5 )
# legend for ratio plot
p_stat = mpatches.Patch(facecolor='0.75', label='Stat.',alpha = 0.5, edgecolor='black' )
p_stat_and_syst = mpatches.Patch(facecolor='yellow', label=r'Stat. $\oplus$ Syst.', alpha = 0.5, edgecolor='black' )
l1 = ax1.legend(handles = [p_stat], loc = 'upper left',
frameon = False, prop = {'size':26})
ax1.legend(handles = [p_stat_and_syst], loc = 'lower left',
frameon = False, prop = {'size':26})
ax1.add_artist(l1)
if variable == 'MET':
ax1.set_ylim( ymin = 0.7, ymax = 1.3 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
# ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
if variable == 'MT':
ax1.set_ylim( ymin = 0.8, ymax = 1.2 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'HT' or variable == 'ST':
ax1.set_ylim( ymin = 0.5, ymax = 1.5 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'WPT':
ax1.set_ylim( ymin = 0.75, ymax = 1.5 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.25 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.05 ) )
if CMS.tight_layout:
plt.tight_layout()
path = output_folder + str( measurement_config.centre_of_mass_energy ) + 'TeV/' + variable + '/' + category
make_folder_if_not_exists( path )
for output_format in output_formats:
filename = path + '/' + histname + '_kv' + str( k_values[channel] ) + '.' + output_format
if channel == 'combined':
filename = filename.replace( '_kv' + str( k_values[channel] ), '' )
plt.savefig( filename )
del hist_data, hist_measured
plt.close()
gc.collect()
htot_usyst = h_bkg_s.Clone()
htot_dsyst = h_bkg_s.Clone()
for ibin in range(1, h_bkg_s.GetNbinsX() + 1):
htot_usyst.SetBinContent(
ibin,
h_bkg_s.GetBinContent(ibin) + h_bkg_s.GetBinErrorUp(ibin))
htot_dsyst.SetBinContent(
ibin,
h_bkg_s.GetBinContent(ibin) - h_bkg_s.GetBinErrorLow(ibin))
rplt.fill_between(htot_usyst,
htot_dsyst,
hatch=r"////////",
alpha=1.0,
linewidth=0,
facecolor="none",
edgecolor="black",
zorder=10,
label=r"Bkgd. Unc")
plot_unc = mpatches.Patch(facecolor="none",
edgecolor="black",
label="Bkgd. Unc",
hatch="////////")
plt.legend(loc=1,
handles=[plot_data, plot_sig1, plot_sig2, plot_bkg, plot_unc],
numpoints=1,
frameon=False)
a2 = plt.axes([0.0, 0.0, 1.0, 0.20], sharex=a1)
def make_plots( histograms, category, output_folder, histname, show_ratio = True, show_before_unfolding = False ):
global variable, phase_space
channel = 'electron'
if 'electron' in histname:
channel = 'electron'
elif 'muon' in histname:
channel = 'muon'
else:
channel = 'combined'
# plot with matplotlib
hist_data = histograms['unfolded']
if category == 'central':
hist_data_with_systematics = histograms['unfolded_with_systematics']
hist_measured = histograms['measured']
hist_data.markersize = 2
hist_data.marker = 'o'
if category == 'central':
hist_data_with_systematics.markersize = 2
hist_data_with_systematics.marker = 'o'
hist_measured.markersize = 2
hist_measured.marker = 'o'
hist_measured.color = 'red'
plt.figure( figsize = CMS.figsize, dpi = CMS.dpi, facecolor = CMS.facecolor )
if show_ratio:
gs = gridspec.GridSpec( 2, 1, height_ratios = [5, 1] )
axes = plt.subplot( gs[0] )
else:
axes = plt.axes()
if variable in ['NJets', 'abs_lepton_eta', 'lepton_eta']:
plt.xlabel( '$%s$' % variables_latex[variable], CMS.x_axis_title )
else:
plt.xlabel( '$%s$ [GeV]' % variables_latex[variable], CMS.x_axis_title )
if not variable in ['NJets']:
axes.minorticks_on()
if variable in ['NJets', 'abs_lepton_eta', 'lepton_eta']:
plt.ylabel( r'$\frac{1}{\sigma} \frac{d\sigma}{d' + variables_latex[variable] + '}$', CMS.y_axis_title )
else:
plt.ylabel( r'$\frac{1}{\sigma} \frac{d\sigma}{d' + variables_latex[variable] + '} \left[\mathrm{GeV}^{-1}\\right]$', CMS.y_axis_title )
plt.tick_params( **CMS.axis_label_major )
if not variable in ['NJets']:
plt.tick_params( **CMS.axis_label_minor )
hist_data.visible = True
if category == 'central':
hist_data_with_systematics.visible = True
rplt.errorbar( hist_data_with_systematics, axes = axes, label = 'do_not_show', xerr = None, capsize = 0, elinewidth = 2, zorder = len( histograms ) + 1 )
rplt.errorbar( hist_data, axes = axes, label = 'do_not_show', xerr = None, capsize = 15, capthick = 3, elinewidth = 2, zorder = len( histograms ) + 2 )
rplt.errorbar( hist_data, axes = axes, label = 'data', xerr = None, yerr = False, zorder = len( histograms ) + 3 ) # this makes a nicer legend entry
if show_before_unfolding:
rplt.errorbar( hist_measured, axes = axes, label = 'data (before unfolding)', xerr = None, zorder = len( histograms ) )
dashes = {}
for key, hist in sorted( histograms.items() ):
zorder = sorted( histograms, reverse = False ).index( key )
if key == 'powhegPythia8' and zorder != len(histograms) - 3:
zorder = len(histograms) - 3
elif key != 'powhegPythia8' and not 'unfolded' in key:
while zorder >= len(histograms) - 3:
zorder = zorder - 1
if not 'unfolded' in key and not 'measured' in key:
hist.linewidth = 4
# setting colours
linestyle = None
if 'amcatnlo_HERWIG' in key or 'massdown' in key:
hist.SetLineColor( kBlue )
dashes[key] = [25,5,5,5,5,5,5,5]
elif 'madgraphMLM' in key or 'scaledown' in key:
hist.SetLineColor( 417 )
dashes[key] = [5,5]
elif 'MADGRAPH_ptreweight' in key:
hist.SetLineColor( kBlack )
elif 'powhegPythia8' in key:
linestyle = 'solid'
dashes[key] = None
hist.SetLineColor( 633 )
elif 'massup' in key or 'amcatnlo' in key:
hist.SetLineColor( 807 )
dashes[key] = [20,5]
elif 'MCATNLO' in key or 'scaleup' in key:
hist.SetLineColor( 619 )
dashes[key] = [5,5,10,5]
if linestyle != None:
hist.linestyle = linestyle
line, h = rplt.hist( hist, axes = axes, label = measurements_latex[key], zorder = zorder )
if dashes[key] != None:
line.set_dashes(dashes[key])
h.set_dashes(dashes[key])
handles, labels = axes.get_legend_handles_labels()
# making data first in the list
data_label_index = labels.index( 'data' )
data_handle = handles[data_label_index]
labels.remove( 'data' )
handles.remove( data_handle )
labels.insert( 0, 'data' )
handles.insert( 0, data_handle )
new_handles, new_labels = [], []
zipped = dict( zip( labels, handles ) )
labelOrder = ['data',
measurements_latex['powhegPythia8'],
measurements_latex['amcatnlo'],
measurements_latex['amcatnlo_HERWIG'],
measurements_latex['madgraphMLM'],
measurements_latex['scaleup'],
measurements_latex['scaledown'],
measurements_latex['massup'],
measurements_latex['massdown']
]
for label in labelOrder:
if label in labels:
new_handles.append(zipped[label])
new_labels.append(label)
legend_location = (0.97, 0.82)
if variable == 'MT':
legend_location = (0.05, 0.82)
elif variable == 'ST':
legend_location = (0.97, 0.82)
elif variable == 'WPT':
legend_location = (1.0, 0.84)
elif variable == 'abs_lepton_eta':
legend_location = (1.0, 0.94)
plt.legend( new_handles, new_labels, numpoints = 1, prop = CMS.legend_properties, frameon = False, bbox_to_anchor=legend_location,
bbox_transform=plt.gcf().transFigure )
label, channel_label = get_cms_labels( channel )
# title
plt.title( label,loc='right', **CMS.title )
# CMS text
# note: fontweight/weight does not change anything as we use Latex text!!!
logo_location = (0.05, 0.98)
prelim_location = (0.05, 0.92)
channel_location = ( 0.05, 0.86)
if variable == 'WPT':
logo_location = (0.03, 0.98)
prelim_location = (0.03, 0.92)
channel_location = (0.03, 0.86)
elif variable == 'abs_lepton_eta':
logo_location = (0.03, 0.98)
prelim_location = (0.03, 0.92)
channel_location = (0.03, 0.86)
plt.text(logo_location[0], logo_location[1], r"\textbf{CMS}", transform=axes.transAxes, fontsize=42,
verticalalignment='top',horizontalalignment='left')
# preliminary
plt.text(prelim_location[0], prelim_location[1], r"\emph{Preliminary}",
transform=axes.transAxes, fontsize=42,
verticalalignment='top',horizontalalignment='left')
# channel text
plt.text(channel_location[0], channel_location[1], r"\emph{%s}" %channel_label, transform=axes.transAxes, fontsize=40,
verticalalignment='top',horizontalalignment='left')
ylim = axes.get_ylim()
if ylim[0] < 0:
axes.set_ylim( ymin = 0.)
if variable == 'WPT':
axes.set_ylim(ymax = ylim[1]*1.3)
elif variable == 'abs_lepton_eta':
axes.set_ylim(ymax = ylim[1]*1.3)
else :
axes.set_ylim(ymax = ylim[1]*1.2)
if show_ratio:
plt.setp( axes.get_xticklabels(), visible = False )
ax1 = plt.subplot( gs[1] )
if not variable in ['NJets']:
ax1.minorticks_on()
#ax1.grid( True, 'major', linewidth = 1 )
# setting the x_limits identical to the main plot
x_limits = axes.get_xlim()
ax1.set_xlim(x_limits)
ax1.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
if not variable in ['NJets']:
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
if variable in ['NJets', 'abs_lepton_eta', 'lepton_eta']:
plt.xlabel('$%s$' % variables_latex[variable], CMS.x_axis_title )
else:
plt.xlabel( '$%s$ [GeV]' % variables_latex[variable], CMS.x_axis_title )
plt.tick_params( **CMS.axis_label_major )
if not variable in ['NJets']:
plt.tick_params( **CMS.axis_label_minor )
plt.ylabel( '$\\frac{\\textrm{pred.}}{\\textrm{data}}$', CMS.y_axis_title )
ax1.yaxis.set_label_coords(-0.115, 0.8)
#draw a horizontal line at y=1 for data
plt.axhline(y = 1, color = 'black', linewidth = 2)
for key, hist in sorted( histograms.iteritems() ):
if not 'unfolded' in key and not 'measured' in key:
ratio = hist.Clone()
ratio.Divide( hist_data ) #divide by data
line, h = rplt.hist( ratio, axes = ax1, label = 'do_not_show' )
if dashes[key] != None:
h.set_dashes(dashes[key])
stat_lower = hist_data.Clone()
stat_upper = hist_data.Clone()
syst_lower = hist_data.Clone()
syst_upper = hist_data.Clone()
# plot error bands on data in the ratio plot
stat_errors = graph_to_value_errors_tuplelist(hist_data)
if category == 'central':
syst_errors = graph_to_value_errors_tuplelist(hist_data_with_systematics)
for bin_i in range( 1, hist_data.GetNbinsX() + 1 ):
stat_value, stat_error, _ = stat_errors[bin_i-1]
stat_rel_error = stat_error/stat_value
stat_lower.SetBinContent( bin_i, 1 - stat_rel_error )
stat_upper.SetBinContent( bin_i, 1 + stat_rel_error )
if category == 'central':
syst_value, syst_error_down, syst_error_up = syst_errors[bin_i-1]
syst_rel_error_down = syst_error_down/syst_value
syst_rel_error_up = syst_error_up/syst_value
syst_lower.SetBinContent( bin_i, 1 - syst_rel_error_down )
syst_upper.SetBinContent( bin_i, 1 + syst_rel_error_up )
if category == 'central':
rplt.fill_between( syst_lower, syst_upper, ax1,
color = 'yellow' )
rplt.fill_between( stat_upper, stat_lower, ax1, color = '0.75',
)
loc = 'upper left'
# if variable in ['ST']:
# loc = 'upper right'
# legend for ratio plot
p_stat = mpatches.Patch(facecolor='0.75', label='Stat.', edgecolor='black' )
p_stat_and_syst = mpatches.Patch(facecolor='yellow', label=r'Stat. $\oplus$ Syst.', edgecolor='black' )
l1 = ax1.legend(handles = [p_stat, p_stat_and_syst], loc = loc,
frameon = False, prop = {'size':26}, ncol = 2)
# ax1.legend(handles = [p_stat_and_syst], loc = 'lower left',
# frameon = False, prop = {'size':30})
ax1.add_artist(l1)
if variable == 'MET':
ax1.set_ylim( ymin = 0.8, ymax = 1.2 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
# ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
if variable == 'MT':
ax1.set_ylim( ymin = 0.8, ymax = 1.2 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'HT':
ax1.set_ylim( ymin = 0.8, ymax = 1.37 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'ST':
ax1.set_ylim( ymin = 0.7, ymax = 1.5 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'WPT':
ax1.set_ylim( ymin = 0.8, ymax = 1.2 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.5 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'NJets':
ax1.set_ylim( ymin = 0.7, ymax = 1.5 )
elif variable == 'abs_lepton_eta':
ax1.set_ylim( ymin = 0.8, ymax = 1.2 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
elif variable == 'lepton_pt':
ax1.set_ylim( ymin = 0.8, ymax = 1.3 )
ax1.yaxis.set_major_locator( MultipleLocator( 0.2 ) )
ax1.yaxis.set_minor_locator( MultipleLocator( 0.1 ) )
if CMS.tight_layout:
plt.tight_layout()
path = '{output_folder}/{centre_of_mass_energy}TeV/{phaseSpace}/{variable}/'
path = path.format(
output_folder = output_folder,
centre_of_mass_energy = measurement_config.centre_of_mass_energy,
phaseSpace = phase_space,
variable = variable
)
make_folder_if_not_exists( path )
for output_format in output_formats:
filename = path + '/' + histname + '.' + output_format
plt.savefig( filename )
del hist_data, hist_measured
plt.close()
gc.collect()
import matplotlib
import matplotlib.pyplot as plt
import numpy as np
import rootpy.plotting.root2matplotlib as rplt
from rootpy.plotting import Hist
x = np.linspace(0, 2 * np.pi, 100)
plt.figure(figsize=(8, 8), dpi=100)
#plt.fill(x,np.sin(x),color='blue',alpha=0.5)
plt.fill(x, np.sin(x), color='None', edgecolor='blue', hatch='///')
plt.fill(x, np.sin(2 * x), color='None', linewidth=0, hatch=r'\\', zorder=950)
a = Hist(5, x[0], x[-1])
b = a.Clone()
b.Fill(3, .5)
rplt.fill_between(
a,
b,
edgecolor='black',
linewidth=0,
facecolor=(0, 0, 0, 0),
hatch=r'\\\\\\\\',
zorder=900,
)
plt.savefig('./test.eps')
plt.savefig('./test.pdf')
import matplotlib
import matplotlib.pyplot as plt
import numpy as np
import rootpy.plotting.root2matplotlib as rplt
from rootpy.plotting import Hist
x = np.linspace(0,2*np.pi,100)
plt.figure(figsize=(8, 8), dpi=100)
#plt.fill(x,np.sin(x),color='blue',alpha=0.5)
plt.fill(x,np.sin(x),color='None', edgecolor='blue', hatch='///')
plt.fill(x,np.sin(2 * x),color='None',linewidth=0,hatch=r'\\', zorder=950)
a = Hist(5, x[0], x[-1])
b = a.Clone()
b.Fill(3, .5)
rplt.fill_between(a, b,
edgecolor='black',
linewidth=0,
facecolor=(0,0,0,0),
hatch=r'\\\\\\\\',
zorder=900,
)
plt.savefig('./test.eps')
plt.savefig('./test.pdf')