return r
#A list of the fit result files, separated by cut and channel
fit_files = [
('final_2j1t_mva', 'ele', os.environ["STPOL_DIR"] +
'/final_fit/results/ele__mva_BDT_with_top_mass_C_eta_lj_el_pt_mt_el_pt_bj_mass_bj_met_mass_lj.txt'
),
('final_2j1t_mva', 'mu', os.environ["STPOL_DIR"] +
'/final_fit/results/mu__mva_BDT_with_top_mass_eta_lj_C_mu_pt_mt_mu_met_mass_bj_pt_bj_mass_lj.txt'
),
]
fitpars = NestedDict()
fitpars_process = NestedDict()
#Load all the files in two formats
for (cut, chan, inf) in fit_files:
fitpars[cut][chan] = from_file(inf, match_fmt='sample')
fitpars_process[cut][chan] = from_file(inf, match_fmt='process')
#Separate SF-s for MET-variated templates, probably no longer needed.
names = [
"t-channel", "top+QCD", "W, diboson"
]
for met in [30, 50, 70]:
met = str(met)
fitpars['final_2j1t_mva_met'+met]['mu'] = from_file(os.environ["STPOL_DIR"]+"/final_fit/results/histos/met%s/mu__mva_BDT_with_top_mass_eta_lj_C_mu_pt_mt_mu_met_mass_bj_pt_bj_mass_lj.txt" % met)
# FIXME -> Currently no MET cut dependent results for electron channel
def combine_templates(templates, patterns, conf):
"""
Args:
"""
hists = {}
hsources = []
for k in ["data", "mc_nom", "mc_varsamp", "mc_varproc"]:
items = templates[patterns[k]]
if len(items)==0:
raise MatchingException("Nothing matched to %s:%s" % (k, patterns[k]))
hsources += items
if len(hsources)==0:
raise ValueError("No histograms matched")
hqcd = NestedDict()
hqcd["nominal"][None] = []
for syst in ["yield", "iso"]:
for sdir in ["up", "down"]:
hqcd[syst][sdir] = []
templates_qcd = templates[patterns["data_antiiso"]]
if len(templates_qcd)==0:
raise MatchingException("Nothing matched to %s:%s" % ("data_antiiso", patterns["data_antiiso"]))
for keys, hist in templates[patterns["data_antiiso"]]:
if keys[1].startswith("antiiso"):
#We have isolation variations
isodir = keys[1].split("_")[1]
if isodir=="nominal":
hqcd["nominal"][None].append(hist)
hup = hist.Clone()
hdown = hist.Clone()
hup.Scale(qcd_yield_variations[0])
hdown.Scale(qcd_yield_variations[1])
hqcd["yield"]["up"].append(hup)
hqcd["yield"]["down"].append(hdown)
elif isodir in ["up", "down"]:
hqcd["iso"][isodir].append(hist)
else:
raise ValueError("Undefined isolation variation direction: %s" % isodir)
#We only have the nominal QCD shape
elif keys[1]=="weight__unweighted":
hqcd["nominal"][None].append(hist)
hup = hist.Clone()
hdown = hist.Clone()
hup.Scale(qcd_yield_variations[0])
hdown.Scale(qcd_yield_variations[1])
hqcd["yield"]["up"].append(hup)
hqcd["yield"]["down"].append(hdown)
#Placeholders for the isolation variation
for isodir in ["up", "down"]:
h = hist.Clone()
hqcd["iso"][isodir].append(h)
else:
raise Exception("Couldn't parse the QCD pattern: %s" % str(keys))
def map_leaves(di, f, equate=True):
for k, v in di.items():
if isinstance(v, dict):
map_leaves(v, f)
else:
if equate:
di[k] = f(v)
else:
f(v)
return di
#Sum the anti-iso data subsamples
map_leaves(hqcd, lambda li: reduce(lambda x,y: x+y, li))
#Normalize the isolation variations to the nominal
map_leaves(hqcd["iso"],
lambda hi:
hi.Scale(hqcd["nominal"][None].Integral() / hi.Integral()) if hi.Integral()>0 else 0,
equate=False
)
#Add the variated data-driven QCD templates
hsources += [
(("data", "qcd", "weight__unweighted"), hqcd["nominal"][None]),
(("data", "qcd", "weight__qcd_yield_up"), hqcd["yield"]["up"]),
(("data", "qcd", "weight__qcd_yield_down"), hqcd["yield"]["down"]),
(("data", "qcd", "weight__qcd_iso_up"), hqcd["iso"]["up"]),
(("data", "qcd", "weight__qcd_iso_down"), hqcd["iso"]["down"]),
]
#f = open('temp.pickle','wb')
#pickle.dump(hsources, f)
#f.close()
#load the histos from the temporary pickle
#f = open('temp.pickle','rb')
#hsources = pickle.load(f)
syst_scenarios = NestedDict()
for (sample_var, sample, weight_var), hist in hsources:
make_hist(hist)
# if "__ele" in weight_var:
# continue
if ".root" in sample:
sample = sample[:sample.index(".root")]
if "__" in weight_var:
spl = weight_var.split("__")
wn = spl[1]
else:
wn = weight_var
sample_var = sample_var.lower()
wtype = None
wdir = None
stype = None
sdir = None
syst = None
#Nominal weight, look for variated samples
if wn=="nominal":
syst = sample_var
elif wn=="unweighted":
syst="unweighted"
else:
#Variated weight, use only nominal sample or data in case of data-driven shapes
if not (sample_var=="nominal" or sample_var=="data"):
continue
syst = wn
if wn==conf.get_nominal_weight() and sample_var=="nominal":
logger.info("Using %s:%s as nominal sample for %s" % (wn, sample_var, sample))
syst_scenarios[sample]["nominal"][None] = hist
#A systematic scenario which has a separate systematic sample
elif sample_var == "syst":
try:
r = get_syst_from_sample_name(sample)
except Exception as e:
logger.warning("Unhandled systematic: %s" % str(e))
r = None
if not r:
continue
sample, systname, d = r
#sample = map_syst_sample_to_nominal(sample)
syst_scenarios[sample][systname][d] = hist
else:
logger.debug("Systematically variated weight: %s:%s %s" % (wn, sample_var, sample))
systname, d = get_updown(syst)
syst_scenarios[sample][systname][d] = hist
logger.info("histogram W3Jets_exclusive nominal: " + "%f %d" % (
syst_scenarios["W3Jets_exclusive"]["nominal"][None].Integral(),
syst_scenarios["W3Jets_exclusive"]["nominal"][None].GetEntries())
)
######################################
### Save systematics, fill missing ###
######################################
#########
# tchan #
#########
#T_t_ToLeptons mass_up is missing, take the mass down and flip the difference with the nominal
mnomt = syst_scenarios["T_t_ToLeptons"]["nominal"][None].Clone()
mdownt = syst_scenarios["T_t_ToLeptons"]["mass"]["down"].Clone()
mupt = (mnomt+mnomt-mdownt)
syst_scenarios["T_t_ToLeptons"]["mass"]["up"] = mupt
#########
# TTBar #
#########
#TTbar variations are provided for the inclusive only, fill them for the exclusive
nom_ttbar = syst_scenarios["TTJets_FullLept"]["nominal"][None] + syst_scenarios["TTJets_SemiLept"]["nominal"][None]
for syst in ["mass", "ttbar_scale", "ttbar_matching"]:
for sample in ["TTJets_FullLept", "TTJets_SemiLept"]:
for sd in ["up", "down"]:
syst_scenarios[sample][syst][sd] = syst_scenarios[sample]["nominal"][None] * syst_scenarios["TTJets"][syst][sd] / nom_ttbar
syst_scenarios.pop("TTJets")
syst_scenarios = syst_scenarios.as_dict()
#Create the output file
p = os.path.dirname(conf.get_outfile_unmerged())
if not os.path.exists(p):
os.makedirs(p)
of = ROOT.TFile(conf.get_outfile_unmerged() , "RECREATE")
of.cd()
#Get the list of all possible systematic scenarios that we have available
allsysts = get_all_systs(syst_scenarios)
for sampn, h1 in syst_scenarios.items():
#Consider all the possible systematic scenarios
for systname in allsysts:
#If we have it available, fine, use it
if systname in h1.keys():
h2 = h1[systname]
#If not, in case of MC and a non-trivial variation
elif not sampn.startswith("Single") and systname not in ["unweighted", "nominal"]:
#Try to get the unvariated template as a placeholder
h = h1.get("nominal", None)
if not h:
h = h1.get("unweighted", None)
if not h:
raise Exception("Could not get the nominal template for %s:%s" % (sampn, systname))
#Our convention is that even the unvariated template is a dict with a single
#key for the direction of variation, which is 'None'
h = h[None]
#Add placeholder templates
for systdir in ["up", "down"]:
h = h.Clone(hname_encode(conf.varname, sampn, systname, systdir))
set_missing_hist(h)
#Save to file
h.SetDirectory(of)
h.Write()
continue
else:
continue
for systdir, h in h2.items():
if systdir==None and systname=="nominal" or not sample_types.is_mc(sampn):
h = h.Clone(hname_encode(conf.varname, sampn))
elif systdir==None and systname=="unweighted":
h = h.Clone(hname_encode(conf.varname, sampn, "unweighted"))
else:
h = h.Clone(hname_encode(conf.varname, sampn, systname, systdir))
h.SetDirectory(of)
h.Write()
nkeys = len(of.GetListOfKeys())
logger.info("Saved %d histograms to file %s" % (nkeys, of.GetPath()))
# of.Close()
########################
### Load systematics ###
########################
# of_unmerged = File(conf.get_outfile_unmerged())
hists = dict()
# ROOT.gROOT.cd()
for k in of.GetListOfKeys():
hists[k.GetName()] = of.Get(k.GetName())
h = hists[k.GetName()]
#hists[k.GetName()].Rebin(2)
logger.info("Loaded %d histograms from file %s" % (len(hists), of.GetPath()))
#of_unmerged.Close()
########################
### Merge ###
########################
from plots.common.utils import merge_hists, PhysicsProcess
merge_cmds = PhysicsProcess.get_merge_dict(
PhysicsProcess.get_proc_dict(conf.channel)
)
hsysts = NestedDict()
for k, v in hists.items():
spl = split_name(k)
hsysts[spl["type"]][spl["dir"]][spl["sample"]] = v
hsysts = hsysts.as_dict()
p = os.path.dirname(conf.get_outfile_merged())
if not os.path.exists(p):
os.makedirs(p)
of = ROOT.TFile(conf.get_outfile_merged(), "RECREATE")
of.cd()
for syst, h1 in hsysts.items():
if syst in skipped_systs:
continue
for sdir, h2 in h1.items():
hmc = merge_hists(h2, merge_cmds)
for hn, h in hmc.items():
if syst=="nominal" or syst=="unweighted":
h.SetName("__".join([spl["var"], hn]))
else:
h.SetName("__".join([spl["var"], hn, syst, sdir]))
h.SetDirectory(of)
h.Write()
nkeys = len(of.GetListOfKeys())
logger.info("Saved %d histograms to file %s" % (nkeys, of.GetPath()))
of.Close()
hists = load_theta_format(conf.get_outfile_merged())
processes = []
systs = []
for (variable, sample, syst, systdir), v in hists.items_flat():
processes.append(sample)
systs.append(syst)
processes = set(processes)
systs = set(systs)
logger.info("Processes: %s" % processes)
if not processes == set(['diboson', 'schan', 'tWchan', 'TTJets', 'tchan', 'WJets', 'qcd', 'DYJets', 'data']):
raise Exception("Combined file did not contain the necessary processes: %s" % str(processes))
logger.info("Systematic scenarios: %s" % systs)
cut = Cuts.final(2,1)*Cuts.mu
var = "cos_theta"
plot_range=[20, -1, 1]
lumi = lumi_iso["mu"]
plot_args = {"x_label": "cos #theta"}
weights = [
("unw", "1.0"),
("nom", "b_weight_nominal"),
("bc_up", "b_weight_nominal_BCup"),
("bc_down", "b_weight_nominal_BCdown"),
("l_up", "b_weight_nominal_Lup"),
("l_down", "b_weight_nominal_Ldown"),
]
hists = NestedDict()
merge_cmds = {
"ttbar": ["TTJets_.*"],
"tchan": [".*_ToLeptons"],
"wjets": ["W[1-4]Jets.*"]
}
pretty_names = {
"ttbar": "t#bar{t}",
"tchan": "signal (t-channel)",
"wjets": "W",
}
pretty_names_weights = {
"unw": "unweighted",
base = os.path.join(os.environ['STPOL_DIR'], 'qcd_estimation', 'fitted', channel)
fn = base + '/%s_no_MC_subtraction_mt_%s_plus.txt' % (cut, met)
fi = open(fn)
li = fi.readline().strip().split()
sf = float(li[0])
#The yield and its uncertainty
y = float(li[1])
uncy = float(li[2])
#Calculate the uncertainty of the scale factor
unc_sf = uncy/y
if not do_uncertainty:
return sf
else:
return sf, unc_sf * sf
qcdScale = dict()
qcdScale['mu'] = dict()
qcdScale['ele'] = dict()
#Determined as the ratio of the integral of anti-iso data after full selection / full selection minus MVA
#See qcd_scale_factors.ipynb for determination
qcd_cut_SF = NestedDict()
qcd_cut_SF['mva']['loose']['mu'] = 0.114754098361
qcd_cut_SF['mva']['loose']['ele'] = 0.0734908136483
qcd_cut_SF['cutbased']['final']['mu'] = 0.0983606557377
qcd_cut_SF['cutbased']['final']['ele'] = 0.0629921259843
qcd_cut_SF = qcd_cut_SF.as_dict()