def load(self):
# Run default loading
templates.Templates.load(self)
mc_channels = set(["mc", ])
channel.expand(self._channel_config, mc_channels)
for var in ['pt', 'eta', 'phi']:
for flavor in ['b', 'c', 'l']:
key = '%s_%cjet' % (var, flavor)
input = '/BTagEff_Chi2sel/%s_%cJet' % (var, flavor)
for channel_, plot_ in self.plots[input].items():
if "data" == channel_ or "qcd" == channel_:
break
elif channel_ in mc_channels:
if key in self._btag:
self._btag[key].Add(plot_)
self._btag[key+'_btag'].Add(self.plots[input+'_bTag'][channel_])
else:
self._btag[key] = plot_.Clone()
self._btag[key+'_btag'] = self.plots[input+'_bTag'][channel_].Clone()
def load(self):
'''
Load all channels into memory
Child classes may explicitly call this function to load plots
'''
bg_channels = set(["mc", ])
channel.expand(self._channel_config, bg_channels)
for channel_ in self._channels:
ch_loader = self._channel_loader(self._prefix,
verbose=self._verbose)
ch_loader.load(self._channel_config, self._plot_config, channel_,
plot_patterns=self._plot_patterns)
channel_scale_ = (self._channel_scale and
self._channel_scale.get(channel_, None))
if self._verbose and channel_scale_:
print("custom scale", channel_,
"x{0:.3f}".format(channel_scale_))
# Channel plots are loaded. Store plots in the dictionary with
# keys equal to plot name and values are dictionaries with keys
# being the channel and values are plots; scale the plots if scale
# is provided
#
for key, hist in ch_loader.plots.items():
if key not in self.plots:
self._plots[key] = {}
if channel_scale_:
hist.Scale(channel_scale_)
if self._bg_error and channel_ in bg_channels:
error.add(hist, self._bg_error)
self._plots[key][channel_] = hist
del(ch_loader)
def __init__(self, options, args, config):
SignalOverBackground.__init__(self, options, args, config)
self._background_channels = set(["mc", ])
channel.expand(self._channel_config, self._background_channels)
def __init__(self, options, args, config,
channel_loader=loader.ChannelLoader):
self._verbose = (options.verbose if options.verbose
else config["core"]["verbose"])
self._batch_mode = (options.batch if options.batch
else config["core"]["batch"])
if self._batch_mode and not options.save and self._verbose:
print("warning: script is run in batch mode but canvases are not "
"saved")
# load channel configuration
#
channel_config = (options.channel_config if options.channel_config
else config["template"]["channel"])
if not channel_config:
raise RuntimeError("channel config is not defined")
self._channel_config = channel.load(os.path.expanduser(channel_config))
# load channel scale(s)
#
self._channel_scale = (
scale.load(os.path.expanduser(options.channel_scale),
self._channel_config)
if options.channel_scale
else None)
# load plot configuration
#
plot_config = (options.plot_config if options.plot_config
else config["template"]["plot"])
if not plot_config:
raise RuntimeError("plot config is not defined")
self._plot_config = plot.load(os.path.expanduser(plot_config))
# Get list of channels to be loaded
#
if not options.channels:
raise RuntimeError("no channels are specified")
use, ban = split_use_and_ban(ch.strip()
for ch in options.channels.split(','))
if not use:
raise RuntimeError("no channels are turned ON")
use = channel.expand(self._channel_config, use, verbose=self._verbose)
if ban:
use -= channel.expand(self._channel_config, ban,
verbose=self._verbose)
if not use:
raise RuntimeError("all channels are turned OFF")
self._channels = use
# Loaded plots are kept in the structures like:
# plot_name: {
# channel1: hist_obj,
# channel2: hist_obj,
# etc.
# }
#
self._plots = {}
self._channel_loader = channel_loader
self._prefix = options.prefix
if options.save:
value = options.save.lower()
if value not in ["ps", "pdf"]:
raise RuntimeError("unsupported save format: " + value)
self._save = value
else:
self._save = False
self._bg_error = options.bg_error
if self._bg_error:
self._bg_error = float(self._bg_error)
if not (0 <= self._bg_error <= 100):
raise RuntimeError("background error is out of range")
else:
self._bg_error /= 100
self._label = options.label or os.getenv("EXO_PLOT_LABEL", None)
self._sub_label = options.sub_label or os.getenv("EXO_PLOT_SUBLABEL",
None)
if options.plots:
self._plot_patterns = options.plots.split(':')
else:
self._plot_patterns = []
self._legend_align = "right"
self._legend_valign = "top"
self._log = options.log
self._ratio= options.ratio
def plot(self):
'''
Process loaded histograms and draw them
The plot() method is responsible for processing loaded channels, e.g.
put data into Stack, conbine signals, etc.
'''
canvases = []
bg_channels = set(["mc", "qcd"])
channel.expand(self._channel_config, bg_channels)
for plot_, channels in self.plots.items():
# Prepare stacks for data, background and signal
signal = ROOT.THStack()
background = ROOT.THStack()
data = None
legend = ROOT.TLegend(0, 0, 0, 0) # coordinates will be adjusted
# prepare channels order and append any missing channels to the
# end in random order
order = self._channel_config["order"]
order.extend(set(channels.keys()) - set(order))
# process channels in order
backgrounds = [] # backgrounds should be added to THStack in
# reverse order to match legend order
for channel_ in order:
if channel_ not in channels:
continue
hist = channels[channel_]
if channel_ in bg_channels:
backgrounds.append(hist)
label = "fe"
elif (channel_.startswith("zp") or
channel_.startswith("rsg")):
# Signal order does not matter
signal.Add(hist)
label = "l"
elif channel_ == "data":
data = hist
label = "lpe"
legend.AddEntry(hist,
self._channel_config["channel"][channel_]
["legend"],
label)
# Add backgrounds to the Stack
if backgrounds:
map(background.Add, reversed(backgrounds))
canvas = self.draw_canvas(plot_,
signal=signal, background=background,
data=data, legend=legend)
if canvas:
canvases.append(canvas)
return canvases
def load(self):
# Run default loading
templates.Templates.load(self)
# Run TFraction Fitter to get MC and QCD scales
try:
if self._verbose:
print("{0:-<80}".format("-- TFractionFitter "))
if self._tff_input not in self.plots:
raise RuntimeError("load plot " + self._tff_input)
# Extract met DATA, QCD and MC
met = {}
mc_channels = set(["mc"])
channel.expand(self._channel_config, mc_channels)
for channel_, plot_ in self.plots[self._tff_input].items():
if "data" == channel_:
met["data"] = plot_
elif "qcd" == channel_:
met["qcd"] = plot_
elif channel_ in mc_channels:
if "mc" in met:
met["mc"].Add(plot_)
else:
met["mc"] = plot_.Clone()
print(met.keys())
missing_channels = set(["data", "qcd", "mc"]) - set(met.keys())
if missing_channels:
raise RuntimeError("channels {0!r} are not loaded".format(missing_channels))
# prepare variable tempaltes for TFractionFitter
templates_ = ROOT.TObjArray(2)
templates_.Add(met["mc"])
templates_.Add(met["qcd"])
# Setup TFractionFitter
fitter = ROOT.TFractionFitter(met["data"], templates_)
# Run TFRactionFitter
fit_status = fitter.Fit()
if fit_status:
raise RuntimeError("fitter error {0}".format(fit_status))
# Extract MC and QCD scales from TFractionFitter and keep
# only central values (drop errors)
fraction = ROOT.Double(0)
fraction_error = ROOT.Double(0)
fractions = {}
fraction_errors = {}
scales = {}
scale_errors = {}
data_integral_ = met["data"].Integral(0, met["data"].GetNbinsX() + 1)
qcd_integral_ = met["qcd"].Integral(0, met["qcd"].GetNbinsX() + 1)
mc_integral_ = met["mc"].Integral(0, met["mc"].GetNbinsX() + 1)
mc_integral_error_ = 0.0
for i in range(0, met["mc"].GetNbinsX() + 2):
mc_integral_error_ = mc_integral_error_ + met["mc"].GetBinError(i) ** 2
mc_integral_error_ = math.sqrt(mc_integral_error_) / mc_integral_
fitter.GetResult(0, fraction, fraction_error)
fractions["mc"] = float(fraction)
fraction_errors["mc"] = float(fraction_error) / float(fraction)
scales["mc"] = float(fraction) * data_integral_ / mc_integral_
scale_errors["mc"] = math.sqrt((1 / data_integral_) + mc_integral_error_ ** 2 + fraction_errors["mc"] ** 2)
fitter.GetResult(1, fraction, fraction_error)
fractions["qcd"] = float(fraction)
fraction_errors["qcd"] = float(fraction_error) / float(fraction)
scales["qcd"] = float(fraction) * data_integral_ / qcd_integral_
scale_errors["qcd"] = math.sqrt(
(1 / data_integral_) + mc_integral_error_ ** 2 + fraction_errors["qcd"] ** 2
)
# Print found fractions
if self._verbose:
print(
"\n".join(
"{0:>3} fraction: {1:.3f} +- {2:.1f}%".format(key.upper(), value, 100 * fraction_errors[key])
for key, value in fractions.items()
)
)
print(
"\n".join(
"{0:>3} scale: {1:.3f} +- {2:.1f}%".format(key.upper(), value, 100 * scale_errors[key])
for key, value in scales.items()
)
)
# Scale all MC and QCD samples with fractions
for plot_, channels_ in self.plots.items():
# Skip normalization if one of the channels is missing
if "data" not in channels_ or "qcd" not in channels_:
continue
# Get MC sum
mc_sum_ = None
for channel_, hist_ in channels_.items():
if channel_ in mc_channels:
if mc_sum_:
mc_sum_.Add(hist_)
else:
mc_sum_ = hist_.Clone()
if not mc_sum_:
continue
data_integral_ = channels_["data"].Integral()
for channel_, hist_ in channels_.items():
if channel_ in mc_channels:
hist_.Scale(scales["mc"])
elif "qcd" == channel_:
hist_.Scale(scales["qcd"])
except RuntimeError as error:
if self._verbose:
print("failed to use TFractionFitter - {0}".format(error), file=sys.stderr)
finally:
if self._verbose:
print()
def plot(self):
""" Process loaded histograms and draw these """
# the main executable script should make sure only one plot is loaded:
# /cutflow or /cutflow_no_weight
channels = self.plots[self.plots.keys().pop()]
signal_channels = set(["zp", "zpwide", "kk"])
channel.expand(self._channel_config, signal_channels)
background_channels = set(["mc"])
channel.expand(self._channel_config, background_channels)
signal_ = {}
background_ = {}
total_background_ = None
data_ = None
for channel_, hist in channels.items():
if channel_ in signal_channels:
signal_[channel_] = cutflow(hist)
elif channel_ in background_channels:
background_[channel_] = cutflow(hist)
if not total_background_:
total_background_ = hist.Clone()
else:
total_background_.Add(hist)
elif channel_ == "data":
data_ = cutflow(hist)
total_background_ = cutflow(total_background_)
channel_names = (
{
"zprime_m1000_w10": r"Z' 1 Tev/c\textsuperscript{2}",
"zprime_m2000_w20": r"Z' 2 Tev/c\textsuperscript{2}",
"zprime_m3000_w30": r"Z' 3 Tev/c\textsuperscript{2}",
"stop": r"Single-Top",
"zjets": r"$Z/\gamma^{\ast}\rightarrow l^{+}l^{-}$",
# "wjets": r"$W\rightarrow l\nu$",
"wb": r"$W\rightarrow l\nu$ (bX)",
"wc": r"$W\rightarrow l\nu$ (cX)",
"wlight": r"$W\rightarrow l\nu$ (lightX)",
"ttbar": r"QCD $t\bar{t}$",
}
if "text" != self._print_mode
else {
"zprime_m1000_w10": r"Z' 1 Tev",
"zprime_m2000_w20": r"Z' 2 Tev",
"zprime_m3000_w30": r"Z' 3 Tev",
"stop": r"Single-Top",
"zjets": r"Z/gamma -> l+ l-",
# "wjets": r"W -> l nu",
"wb": r"W-> l nu (bX)",
"wc": r"W-> l nu (cX)",
"wlight": r"W-> l nu (lightX)",
"ttbar": r"QCD ttbar",
}
)
print_function = print_cutflow_in_text if self._print_mode == "text" else print_cutflow_in_tex
fields_to_print = [["jets", "electron", "veto_lepton", "twod_cut"], ["jet1", "htlep", "tricut", "met"]]
if self._non_threshold:
fields_to_print.append(["chi2", "non_threshold", "eff"])
for samples_cutflow_ in (signal_, background_):
if not samples_cutflow_:
continue
for channel_, cutflow_ in samples_cutflow_.items():
cutflow_["eff"] = efficiency(cutflow_.get("non_threshold", 0), cutflow_.get("chi2", 0))
for cutflow_ in (total_background_, data_):
cutflow_["eff"] = efficiency(cutflow_.get("non_threshold", 0), cutflow_.get("chi2", 0))
for fields in fields_to_print:
print_function(None, None, fields)
for channel_ in self._channel_config["order"]:
if not channel_.startswith("zprime") or channel_ not in signal_:
continue
print_function(channel_names.get(channel_, channel_), signal_[channel_], fields)
for channel_ in ["stop", "zjets", "wb", "wc", "wlight", "ttbar"]:
if channel_ not in background_:
continue
print_function(channel_names.get(channel_, channel_), background_[channel_], fields)
print_function("Total MC", total_background_, fields)
if data_:
print_function("Data 2011", data_, fields)
print()
return None