def shape_variation(var, plot_range, varname, recreate, out_name):
tot_hists = OrderedDict()
tdrstyle()
for wn, w in [
("unw", Weight("1.0")),
("nominal", Weights.wjets_madgraph_shape_weight("nominal")),
("shape_up", Weights.wjets_madgraph_shape_weight("wjets_up")),
("shape_down", Weights.wjets_madgraph_shape_weight("wjets_down"))
]:
p = data_mc(
var, "2J0T_%s" % (wn),
cut, Weights.total("mu", "nominal")*w,
samples, ".", recreate, lumi_iso["mu"], plot_range=plot_range,
)
for hn, h in p.hists.items():
logger.debug("%s %.2f" % (hn, h.Integral()))
#sumw = [(sample.name, numpy.mean(root_numpy.root2array(sample.tfile.GetPath()[:-2], "trees/WJets_weights", branches=[str(w)])[str(w)])) for sample in samples]
s = sum(p.hists.values()).Clone()
tot_hists[wn] = s
hc = HistCollection(tot_hists, name=out_name)
for hn, h in hc.hists.items():
print hn, h.Integral()
canv = plot_hists_dict(hc.hists, do_chi2=False, do_ks=True, x_label=varname, legend_pos="top-left")
hc.hists.values()[0].SetTitle("shape variation")
canv.SaveAs(out_name + ".png")
hc.save(out_name)
return hc, canv
tree = args.tree
#The enabled systematic up/down variation sample prefixes to put on the ratio plot
if args.do_systs:
systs = {
"JES": ("EnUp", "EnDown"),
"JER": ["ResUp", "ResDown"],
"MET": ["UnclusteredEnUp", "UnclusteredEnDown"],
}
else:
systs = {}
for lepton_channel in args.channels:
logger.info("Plotting lepton channel %s" % lepton_channel)
weight = Weights.total(lepton_channel)*Weights.wjets_madgraph_shape_weight()*Weights.wjets_madgraph_flat_weight()
physics_processes = PhysicsProcess.get_proc_dict(lepton_channel=lepton_channel)#Contains the information about merging samples and proper pretty names for samples
merge_cmds = PhysicsProcess.get_merge_dict(physics_processes) #The actual merge dictionary
lumi = lumis[lepton_channel]
isoreg='iso'
use_antiiso = False
if args.use_antiiso:
isoreg='antiiso'
use_antiiso = True
#Get the file lists
flist = get_file_list(
merge_cmds,
from plots.common.sample import Sample
from plots.common.sample_style import ColorStyleGen
from plots.common.hist_plots import plot_hists
from plots.common.cuts import Cuts, Weights
from plots.common.legend import legend
from plots.common.tdrstyle import tdrstyle
if __name__=="__main__":
samp = Sample.fromFile("~/Documents/stpol/data/out_step3_joosep_11_07_19_44/mu/iso/nominal/W4Jets_exclusive.root")
samp.tree.AddFriend("trees/WJets_weights", samp.file_name)
tdrstyle()
cut = str(Cuts.final(2,0)*Cuts.Wflavour("W_heavy"))
mean_weight = samp.drawHistogram(str(Weights.wjets_madgraph_weight("nominal")), cut, weight="1.0", plot_range=[200, 0, 2]).hist.GetMean()
print "mean weight=%.2f" % mean_weight
hi0 = samp.drawHistogram("cos_theta", cut, weight="1.0", plot_range=[20, -1, 1]).hist
hi0.Scale(samp.lumiScaleFactor(20000))
hi1 = samp.drawHistogram("cos_theta", cut, weight=str(Weights.wjets_madgraph_weight("nominal")), plot_range=[20, -1, 1]).hist
hi1.Scale(samp.lumiScaleFactor(20000))
hi2 = samp.drawHistogram("cos_theta", cut, weight=str(Weights.wjets_madgraph_weight("wjets_up")), plot_range=[20, -1, 1]).hist
hi2.Scale(samp.lumiScaleFactor(20000))
hi3 = samp.drawHistogram("cos_theta", cut, weight=str(Weights.wjets_madgraph_weight("wjets_down")), plot_range=[20, -1, 1]).hist
hi3.Scale(samp.lumiScaleFactor(20000))
hists = [hi0, hi1, hi2, hi3]
#for h in hists:
# h.Scale(1.0/h.Integral())
hi0.SetTitle("unweighted")
hi1.SetTitle("weighted")
hi2.SetTitle("weighted wjets_up")
lumi = lumis[args.lumitag]['iso'][proc]
flist = []
for i in ['sig','bg']:
for j in ['train','eval']:
flist += mvaFileList[i][j]
logger.debug('Used file list: %s',str(flist))
# Read in the samples
samples = {}
for f in flist:
samples[f] = Sample.fromFile(args.indir+'/%s/%s.root' % (proc,f), tree_name='Events_MVA')
# Set the weight expression
weightString = str(Weights.total(proc) *
Weights.wjets_madgraph_shape_weight() *
Weights.wjets_madgraph_flat_weight())
# Which file do we use to write our TMVA trainings
ext=''
if len(rVar):
ext='_sans'
for v in rVar:
ext+='_'+v
if len(cVar):
ext='_with'
for v in cVar:
ext+='_'+v
def plot_sherpa_vs_madgraph(var, cut_name, cut, samples, out_dir, recreate=False, **kwargs):
hname = var["varname"]
out_dir = out_dir + "/" + cut_name
if recreate and os.path.exists(out_dir):
logger.info("Output directory %s exists, removing" % out_dir)
shutil.rmtree(out_dir)
mkdir_p(out_dir)
logger.info("Using output directory %s" % out_dir)
logger.info("Using output directory %s" % out_dir)
coll = data_mc(var["var"], cut_name, cut, Weights.total()*Weights.mu, samples, out_dir, recreate, LUMI_TOTAL, reweight_madgraph=True, flavour_split=True, plot_range=var["range"], **kwargs)
logging.debug(str(coll.hists))
for hn, hist in coll.hists.items():
sample_name = coll.metadata[hn].sample_name
process_name = coll.metadata[hn].process_name
match = re.match(".*/cut__flavour__(W_[Hl][Hl])/.*", hn)
if match:
flavour_scenario = match.group(1)
else:
flavour_scenario = None
try:
if sample_types.is_mc(sample_name):
Styling.mc_style(hist, process_name)
else:
Styling.data_style(hist)
except KeyError as e:
logger.warning("Couldn't style histogram %s" % hn)
if flavour_scenario:
logger.debug("Matched flavour split histogram %s, %s" % (hn, flavour_scenario))
#Styling.mc_style(hist, process_name)
if re.match("W_H[lH]", flavour_scenario):
logger.debug("Changing colour of %s" % (hn))
hist.SetFillColor(hist.GetFillColor()+1)
hist.SetLineColor(hist.GetLineColor()+1)
logger.debug("pre merge: %s" % str([ (hn, coll.hists[hn].GetLineColor()) for hn in coll.hists.keys() if "sherpa" in hn]))
merges = dict()
merge_cmds = get_merge_cmds()
merge_cmds.pop("WJets")
merges["madgraph/unweighted"] = merge_cmds.copy()
merges["madgraph/weighted"] = merge_cmds.copy()
merges["sherpa/unweighted"] = merge_cmds.copy()
merges["sherpa/weighted"] = merge_cmds.copy()
merges["sherpa/unweighted"]["WJets_hf"] = ["weight__nominal/cut__flavour__W_heavy/WJets_sherpa_nominal"]
merges["sherpa/unweighted"]["WJets_lf"] = ["weight__nominal/cut__flavour__W_light/WJets_sherpa_nominal"]
merges["sherpa/weighted"]["WJets_hf"] = ["weight__sherpa_flavour/cut__flavour__W_heavy/WJets_sherpa_nominal"]
merges["sherpa/weighted"]["WJets_lf"] = ["weight__sherpa_flavour/cut__flavour__W_light/WJets_sherpa_nominal"]
merges["madgraph/unweighted"]["WJets_hf"] = ["weight__nominal/cut__flavour__W_heavy/W[1-4]Jets_exclusive"]
merges["madgraph/unweighted"]["WJets_lf"] = ["weight__nominal/cut__flavour__W_light/W[1-4]Jets_exclusive"]
merges["madgraph/weighted"]["WJets_hf"] = ["weight__reweight_madgraph/cut__flavour__W_heavy/W[1-4]Jets_exclusive"]
merges["madgraph/weighted"]["WJets_lf"] = ["weight__reweight_madgraph/cut__flavour__W_light/W[1-4]Jets_exclusive"]
hmerged = dict()
for k in merges.keys():
hmerged[k] = merge_hists(copy.deepcopy(coll.hists), merges[k])
logger.debug("post merge: %s" % str([ (hn, hmerged["sherpa/weighted"][hn].GetLineColor()) for hn in hmerged["sherpa/weighted"].keys()]))
#w_mg_sh = 1.0416259307303726 #sherpa to madgraph ratio
w_mg_sh = 1.0821535639376414
hmerged["sherpa/weighted"]["WJets_hf"].Scale(w_mg_sh)
hmerged["sherpa/weighted"]["WJets_lf"].Scale(w_mg_sh)
logger.info("Drawing madgraph unweighted plot")
canv = ROOT.TCanvas("c2", "c2")
suffix = "__%s__%s" % (var["var"], cut_name)
suffix = escape(suffix)
plot(canv, "madgraph_unw"+suffix, hmerged["madgraph/unweighted"], out_dir, **kwargs)
kwargs = dict({"x_label": var["varname"]}, **kwargs)
for k, v in hmerged.items():
logger.debug("Group %s" % k)
for hn, h in v.items():
logger.debug("Sample %s = %.2f" % (hn, h.Integral()))
logger.info("%s data=%.2f" % (k, v["data"].Integral()))
logger.info("%s MC=%.2f" % (k, sum([h.Integral() for k, h in v.items() if k!="data"])))
hists_flavours_merged = dict()
hists_flavours_merged["madgraph/weighted"] = merge_hists(hmerged["madgraph/weighted"], {"WJets": ["WJets_hf", "WJets_lf"]})
hists_flavours_merged["madgraph/unweighted"] = merge_hists(hmerged["madgraph/unweighted"], {"WJets": ["WJets_hf", "WJets_lf"]})
hists_flavours_merged["sherpa/unweighted"] = merge_hists(hmerged["sherpa/unweighted"], {"WJets": ["WJets_hf", "WJets_lf"]})
hists_flavours_merged["sherpa/weighted"] = merge_hists(hmerged["sherpa/weighted"], {"WJets": ["WJets_hf", "WJets_lf"]})
logger.info("Drawing sherpa weighted plot")
canv = ROOT.TCanvas("c1", "c1")
plot(canv, "sherpa_rew"+suffix, hmerged["sherpa/weighted"], out_dir, **kwargs)
logger.info("Drawing sherpa unweighted plot")
canv = ROOT.TCanvas("c1", "c1")
plot(canv, "sherpa_unw"+suffix, hmerged["sherpa/unweighted"], out_dir, **kwargs)
logger.info("Drawing madgraph plot")
canv = ROOT.TCanvas("c2", "c2")
plot(canv, "madgraph_rew"+suffix, hmerged["madgraph/weighted"], out_dir, **kwargs)
total_madgraph = copy.deepcopy(hmerged["madgraph/unweighted"])
merged_colls = dict()
for k, v in hmerged.items():
merged_colls[k] = HistCollection(copy.deepcopy(v), name=k)
logger.info("Drawing sherpa vs. madgraph shape comparison plots")
hists = [
("sherpa unw hf", hmerged["sherpa/unweighted"]["WJets_hf"]),
("sherpa rew hf", hmerged["sherpa/weighted"]["WJets_hf"]),
("madgraph unw hf", hmerged["madgraph/unweighted"]["WJets_hf"]),
("madgraph rew hf", hmerged["madgraph/weighted"]["WJets_hf"]),
]
hists = copy.deepcopy(hists)
for hn, h in hists:
h.SetTitle(hn + " %.2f" % h.Integral())
h.Scale(1.0/h.Integral())
hists = [h[1] for h in hists]
ColorStyleGen.style_hists(hists)
canv = plot_hists(hists, x_label=var["varname"], do_chi2=True)
leg = legend(hists, styles=["f", "f"], **kwargs)
canv.SaveAs(out_dir + "/weighted_flavour_hf_%s.png" % hname)
canv.Close()
hists = [
("data", hmerged["madgraph/unweighted"]["data"]),
("sherpa", hists_flavours_merged["sherpa/unweighted"]["WJets"]),
("madgraph", hists_flavours_merged["madgraph/unweighted"]["WJets"]),
]
hists = copy.deepcopy(hists)
for hn, h in hists:
h.SetTitle(hn + " %.2f" % h.Integral())
h.Scale(1.0/h.Integral())
hists = [h[1] for h in hists]
ColorStyleGen.style_hists(hists)
canv = plot_hists(hists, x_label=var["varname"], do_chi2=True)
leg = legend(hists, styles=["f", "f"], **kwargs)
canv.SaveAs(out_dir + "/unweighted_sherpa_mg_%s.png" % hname)
canv.Close()
hists = [
("data", hmerged["madgraph/unweighted"]["data"]),
("sherpa", hists_flavours_merged["sherpa/weighted"]["WJets"]),
("madgraph", hists_flavours_merged["madgraph/weighted"]["WJets"]),
]
hists = copy.deepcopy(hists)
for hn, h in hists:
h.SetTitle(hn + " %.2f" % h.Integral())
h.Scale(1.0/h.Integral())
hists = [h[1] for h in hists]
ColorStyleGen.style_hists(hists)
canv = plot_hists(hists, x_label=var["varname"], do_chi2=True)
leg = legend(hists, styles=["f", "f"], **kwargs)
canv.SaveAs(out_dir + "/weighted_sherpa_mg_%s.png" % hname)
canv.Close()
hists = [
("sherpa unw lf", hmerged["sherpa/unweighted"]["WJets_lf"]),
("sherpa rew lf", hmerged["sherpa/weighted"]["WJets_lf"]),
("madgraph unw lf", hmerged["madgraph/unweighted"]["WJets_lf"]),
("madgraph rew lf", hmerged["madgraph/weighted"]["WJets_lf"]),
]
hists = copy.deepcopy(hists)
for hn, h in hists:
h.SetTitle(hn + " %.2f" % h.Integral())
h.Scale(1.0/h.Integral())
hists = [h[1] for h in hists]
ColorStyleGen.style_hists(hists)
canv = plot_hists(hists, x_label=var["varname"], do_chi2=True)
leg = legend(hists, styles=["f", "f"], **kwargs)
canv.SaveAs(out_dir + "/weighted_flavour_lf_%s.png" % hname)
canv.Close()
hists = [
("data", hmerged["madgraph/unweighted"]["data"]),
("madgraph unw", hists_flavours_merged["madgraph/unweighted"]["WJets"]),
("madgraph rew", hists_flavours_merged["madgraph/weighted"]["WJets"]),
("sherpa unw", hists_flavours_merged["sherpa/unweighted"]["WJets"]),
("sherpa rew", hists_flavours_merged["sherpa/weighted"]["WJets"]),
]
hists = copy.deepcopy(hists)
for hn, h in hists:
h.SetTitle(hn + " %.2f" % h.Integral())
h.Scale(1.0/h.Integral())
hists = [h[1] for h in hists]
ColorStyleGen.style_hists(hists)
canv = plot_hists(hists, x_label=var["varname"], do_chi2=True)
leg = legend(hists, styles=["f", "f"], **kwargs)
hists[0].SetTitle("")
canv.Update()
canv.SaveAs(out_dir + "/shapes_%s.png" % hname)
canv.Close()
# hists = [
# ("sherpa hf", hmerged["sherpa"]["WJets_hf"]),
# ("madgraph unw hf", hmerged["madgraph/unweighted"]["WJets_hf"]),
# ("madgraph rew hf", hmerged["madgraph/weighted"]["WJets_hf"]),
# ]
# hists = copy.deepcopy(hists)
# for hn, h in hists:
# h.SetTitle(hn + " %.2f" % h.Integral())
# h.Scale(1.0/h.Integral())
# hists = [h[1] for h in hists]
# ColorStyleGen.style_hists(hists)
# canv = plot_hists(hists, x_label=var["varname"], do_chi2=True)
# leg = legend(hists, styles=["f", "f"], **kwargs)
# hists[0].SetTitle("madgraph sherpa rew hf")
# canv.SaveAs(out_dir + "/shapes_hf_%s.png" % hname)
# canv.Close()
return coll, merged_colls
def plot_ratios(cut_name, cut, samples, out_dir, recreate, flavour_scenario=flavour_scenarios[0]):
out_dir += "/" + cut_name
mkdir_p(out_dir)
colls = dict()
samples_WJets = filter(lambda x: sample_types.is_wjets(x.name), samples)
for sc in flavour_scenario:
logger.info("Drawing ratio with cut %s" % sc)
cut_ = cut*getattr(Cuts, sc)
colls[sc] = data_mc(costheta["var"], cut_name + "__" + sc, cut_, Weights.total()*Weights.mu, samples_WJets, out_dir, recreate, LUMI_TOTAL, plot_range=costheta["range"])
logger.debug(colls[flavour_scenario[0]].hists["weight__nominal/cut__all/WJets_sherpa_nominal"].Integral())
logger.debug(colls[flavour_scenario[1]].hists["weight__nominal/cut__all/WJets_sherpa_nominal"].Integral())
coll = dict()
for k, c in colls.items():
for hn, h in c.hists.items():
coll[hn + "/" + k] = h
for k, h in coll.items():
logger.debug("%s = %s" % (k, str([y for y in h.y()])))
logger.debug(coll)
#coll = HistCollection(coll, name=cut_name)
merges = {}
for sc in flavour_scenario:
merges["madgraph/%s" % sc] = ["weight__nominal/cut__all/W[1-4]Jets_exclusive/%s" % sc]
merges["sherpa/unweighted/%s" % sc] = ["weight__nominal/cut__all/WJets_sherpa_nominal/%s" % sc]
merges["sherpa/weighted/%s" % sc] = ["weight__sherpa_flavour/cut__all/WJets_sherpa_nominal/%s" % sc]
merged = merge_hists(coll, merges)
for k, h in merged.items():
logger.debug("%s = %s" % (k, str([y for y in h.y()])))
hists_flavour = dict()
hists_flavour["madgraph"] = ROOT.TH1F("madgraph", "madgraph", len(flavour_scenario), 0, len(flavour_scenario)-1)
hists_flavour["sherpa/unweighted"] = ROOT.TH1F("sherpa_unw", "sherpa unweighted", len(flavour_scenario), 0, len(flavour_scenario)-1)
hists_flavour["sherpa/weighted"] = ROOT.TH1F("sherpa_rew", "sherpa weighted", len(flavour_scenario), 0, len(flavour_scenario)-1)
for i, sc in zip(range(1,len(flavour_scenario)+1), flavour_scenario):
sh1_int, sh1_err = calc_int_err(merged["sherpa/unweighted/%s" % sc])
sh2_int, sh2_err = calc_int_err(merged["sherpa/weighted/%s" % sc])
mg_int, mg_err = calc_int_err(merged["madgraph/%s" % sc])
logger.debug("%.2f %.2f" % (sh1_int, sh1_err))
logger.debug("%.2f %.2f" % (sh2_int, sh2_err))
logger.debug("%.2f %.2f" % (mg_int, mg_err))
hists_flavour["madgraph"].SetBinContent(i, mg_int)
hists_flavour["madgraph"].SetBinError(i, mg_err)
hists_flavour["sherpa/unweighted"].SetBinContent(i, sh1_int)
hists_flavour["sherpa/unweighted"].SetBinError(i, sh1_err)
hists_flavour["sherpa/weighted"].SetBinContent(i, sh2_int)
hists_flavour["sherpa/weighted"].SetBinError(i, sh2_err)
hists_flavour["madgraph"].GetXaxis().SetBinLabel(i, sc)
hists_flavour["sherpa/unweighted"].GetXaxis().SetBinLabel(i, sc)
hists_flavour["sherpa/weighted"].GetXaxis().SetBinLabel(i, sc)
hists_flavour["sherpa/weighted"].Sumw2()
hists_flavour["sherpa/unweighted"].Sumw2()
hists_flavour["madgraph"].Sumw2()
hists_flavour["ratio/unweighted"] = hists_flavour["madgraph"].Clone("ratio_unw")
hists_flavour["ratio/unweighted"].Divide(hists_flavour["sherpa/unweighted"])
hists_flavour["ratio/weighted"] = hists_flavour["madgraph"].Clone("ratio_rew")
hists_flavour["ratio/weighted"].Divide(hists_flavour["sherpa/weighted"])
for i, sc in zip(range(1,len(flavour_scenario)+1), flavour_scenario):
logger.info("weights[%s] = %.6f; //error=%.6f [%d]" % (sc, hists_flavour["ratio/unweighted"].GetBinContent(i), hists_flavour["ratio/unweighted"].GetBinError(i), i))
flavour_ratio_coll = HistCollection(hists_flavour, name="hists__flavour_ratios")
flavour_ratio_coll.save(out_dir)
for sc in flavour_scenario:
hists = [merged["madgraph/%s" % sc], merged["sherpa/unweighted/%s" % sc], merged["sherpa/weighted/%s" % sc]]
for hist in hists:
norm(hist)
#hist.SetName(sc)
#hist.SetTitle(sc)
ColorStyleGen.style_hists(hists)
canv = plot_hists(hists, x_label=costheta["varname"])
leg = legend(hists, styles=["f", "f"], nudge_x=-0.2)
chi2 = hists[0].Chi2Test(hists[1], "WW CHI2/NDF")
hists[0].SetTitle("madgraph to sherpa comparison #chi^{2}/ndf=%.2f" % chi2)
canv.Update()
canv.SaveAs(out_dir + "/flavours__%s.png" % (sc))
md_merged = dict()
for sc in flavour_scenario:
logger.info("Calculating ratio for %s" % sc)
hi = merged["sherpa/unweighted/%s" % sc].Clone("ratio__%s" % sc)
hi.Divide(merged["madgraph/%s" % sc])
merged[hi.GetName()] = hi
hc_merged = HistCollection(merged, md_merged, "hists__costheta_flavours_merged")
hc_merged.save(out_dir)
logger.info("Saved merged histogram collection")
"var": "phi_met",
"binning": [20, -3.14, 3.14]
}
variables["cos_theta"] = {
"name": "cos_theta",
"var": "cos_theta",
"binning": [20, -1, 1]
}
variables["bdt"] = {
"name": "bdt",
"var": Cuts.mva_vars['mu'],
"binning": [60, -1, 1]
}
weights = [
("weight__nominal", Weights.total_weight("mu")),
]
c1 = (
"2j0t",
Cuts.mt_mu()*Cuts.n_jets(2)*Cuts.n_tags(0)*Cuts.lepton("mu")*Cuts.hlt("mu"),
)
c2 = (
"2j1t",
Cuts.mt_mu()*Cuts.n_jets(2)*Cuts.n_tags(1)*Cuts.lepton("mu")*Cuts.hlt("mu"),
)
c3 = (
"final_cb_2j1t",
Cuts.final(2,1),
)
#Define all the variables that we want to see
variables = [
("cos_theta", "cos_theta", [20, -1, 1]),
#Demonstrate some weird binning
("cos_theta_binned", "cos_theta", [-1, -0.4674, -0.2558, -0.1028, 0.0, 0.116, 0.2126, 0.2956, 0.379, 0.4568, 0.5302, 0.6038, 0.6822, 0.7694, 1]),
#ROOT functions can be used
("abs_eta_lj", "abs(eta_lj)", [20, 0, 5]),
("bdt", Cuts.mva_vars['mu'], [60, -1, 1]),
]
#Define all the weight strategies that we want to apply
weights = [
("weight__nominal", Weights.total_weight("mu")),
#Demonstrate reweighting
("weight__puw", Weights.pu("nominal")),
("weight__puw_up", Weights.pu("up")),
]
#All the cut regions that we are interested in
cuts = [
("2j0t", Cuts.mt_mu()*Cuts.n_jets(2)*Cuts.n_tags(0)*Cuts.lepton("mu")*Cuts.hlt("mu")),
("2j1t", Cuts.mt_mu()*Cuts.n_jets(2)*Cuts.n_tags(1)*Cuts.lepton("mu")*Cuts.hlt("mu")),
("final_cb_2j1t", Cuts.final(2,1))
]
#Construct the analysis chain
snodes, out = analysis_tree(cuts, weights, variables, args.infiles, args.outfile)
