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()
# need to set things like min, max, change to log, etc for hstack and soloHists
normalizeTo = plots_path.get('normalize', plots_config.get('normalize', None))
if normalizeTo is not None:
dataScale = 0.
if "unity" in normalizeTo:
for hist in hstack:
if (hist.integral() != 0):
hist.scale(1.0/hist.integral())
for hist in soloHists:
if (hist.integral() != 0):
hist.scale(1.0/hist.integral())
elif "stack" in normalizeTo:
for hist in soloHists:
if (hist.integral() != 0):
print(hist)
print(hstack.Integral()/hist.integral())
exit()
hist.scale(hstack.Integral()/hist.integral())
else:
if normalizeTo not in [hist.title for hist in soloHists]: raise ValueError("Could not find %s as a solo hist for normalizing to." % normalizeTo)
for hist in soloHists:
if hist.title == normalizeTo: dataScale = hist.integral()
mcScale = 0.
for hist in hstack:
mcScale += hist.integral()
if mcScale != 0.:
normalizeFactor = dataScale/mcScale
else:
normalizeFactor = 1.
for hist in hstack: