def GetfromLHAPDF(self, gen):
if not self.SetupLHAPDF():
return 0, None
## Default PDF used for 2017 and 2018 production
if self.pset is None:
self.pset = lhapdf.getPDFSet(
"NNPDF31_nnlo_as_0118") # ErrorType: replicas
self.pdfs = self.pset.mkPDFs()
npar = self.pset.errorType.count(
"+") # number of parameter variations (alphaS, etc.)
if npar > 0:
print("Last %d members are parameter variations\n" % (2 * npar))
## Fill vectors xgAll and xuAll using all PDF members.
pdfweights = [0.0 for i in range(self.pset.size)]
x1_cen = self.pdfs[0].xfxQ(gen.id1, gen.x1, gen.scalePDF)
x2_cen = self.pdfs[0].xfxQ(gen.id2, gen.x2, gen.scalePDF)
normweight = x1_cen * x2_cen
for imem in range(self.pset.size):
x1_imem = self.pdfs[imem].xfxQ(gen.id1, gen.x1, gen.scalePDF)
x2_imem = self.pdfs[imem].xfxQ(gen.id2, gen.x2, gen.scalePDF)
pdfweights[imem] = x1_imem * x2_imem / normweight
return self.pset.size, pdfweights
def main(
pdfset,
flavours,
output_filename,
n_points,
q,
q2, # q2 instead of q
members,
folder):
"""evaluate a PDF set and write the resuling TGraph to disk"""
if folder is not None and not os.path.exists(folder):
os.makedirs(folder)
out = ROOT.TFile((folder+"/" if folder is not None else "") + output_filename, "RECREATE")
x_values = np.logspace(-4, -0.01, n_points) # TODO get min, max x from PDF set
## Version info, search paths, and metadata
print "LHAPDF version", lhapdf.version()
lhapdf.pathsPrepend(os.getcwd())
lhapdf.setVerbosity(0)
print "LHAPDF paths",lhapdf.paths()
pset = lhapdf.getPDFSet(pdfset)
print pset.description
n_members = pset.size
if members is not None:
n_members=members
# iterate over flavours, get pdfgraph, write
for flavour in flavours:
tgraph = get_pdf_tgraph(pset, flavour, x_values, n_points, n_members, q, q2)
tgraph.Write(partondict[flavour].replace(' ', '_'))
print "Written to", output_filename
out.Close()
def generate_pdf(pdfname):
pdf, pdf_index = [], []
pdfset = lhapdf.getPDFSet(f"{pdfname}")
for i in range(1, pdfset.size):
pdf_index.append(i)
for i in range(0, pdfset.size):
pdf.append(pdfset.mkPDF(i))
return pdf, pdf_index
def generate_random_pdf(pdfname, compression_size):
pdf, pdf_index = [], []
pdfset = lhapdf.getPDFSet(f"{pdfname}")
for i in range(1, compression_size):
pdf_index.append(i)
indices = np.random.choice(pdfset.size, compression_size, replace=False)
for i in range(0, compression_size):
pdf.append(pdfset.mkPDF(indices[i]))
return pdf, pdf_index
def _read_lhapdf(self, lhgrid_filename):
pdf = {}
global pdfs
pdfset = lhapdf.getPDFSet(lhgrid_filename)
if not lhgrid_filename in pdfs:
print 'cache it.'
print pdfs.keys()
pdfs[lhgrid_filename] = pdfset.mkPDFs()
else:
print 'is already cached'
for flavor in self._flavors:
pdf[flavor] = self._get_lhapdf_flavor(flavor, pdfs[lhgrid_filename])
return pdf
def f2p_lhapdf(x, q2, select='CT14nnlo'):
pdf_set = lhapdf.getPDFSet(select)
pdfs = pdf_set.mkPDFs()
d, ed = numpy.empty_like(x), numpy.empty_like(x)
dbar, edbar = numpy.empty_like(x), numpy.empty_like(x)
u, eu = numpy.empty_like(x), numpy.empty_like(x)
ubar, eubar = numpy.empty_like(x), numpy.empty_like(x)
s, es = numpy.empty_like(x), numpy.empty_like(x)
for ix, xx in enumerate(x):
d[ix], ed[ix] = _parton_lhapdf(1, xx, q2, pdf_set, pdfs)
dbar[ix], edbar[ix] = _parton_lhapdf(-1, xx, q2, pdf_set, pdfs)
u[ix], eu[ix] = _parton_lhapdf(2, xx, q2, pdf_set, pdfs)
ubar[ix], eubar[ix] = _parton_lhapdf(-2, xx, q2, pdf_set, pdfs)
s[ix], es[ix] = _parton_lhapdf(3, xx, q2, pdf_set, pdfs)
result = ((2 / 3)**2 * (u + ubar) + (1 / 3)**2 * (d + dbar) + (1 / 3)**2 * 2 * s)
error = numpy.sqrt((2 / 3)**4 * (eu**2 + eubar**2) + (1 / 3)**4 * (ed**2 + edbar**2) + (1 / 3)**4 * 4 * es**2)
return result, error
def generate_gp(prior):
""" Generate the GP from an input `prior` PDF """
# Get PDFSet and replicas (split is to avoid replica 0)
pdfset = lhapdf.getPDFSet(prior)
replicas = pdfset.mkPDFs()[1:]
# Initial scale and x-grid
Q0 = float(pdfset.get_entry("QMin"))
xs, nx = XGRID, len(XGRID)
# Available flavours
flavours = get_active_flavours(pdfset, Q0)
nfl = len(flavours)
print(f"{Fore.GREEN}Sampling {nx} x-points at initial scale: {Q0} GeV")
grid_points = list(itertools.product(flavours, xs))
pdf_values = np.empty([nfl * nx, len(replicas)])
for irep, rep in enumerate(replicas):
for ipt, pt in enumerate(grid_points):
pdf_values[ipt][irep] = rep.xfxQ(pt[0], pt[1], Q0)
print("Computing stats")
mean = np.mean(pdf_values, axis=1)
covariance = np.cov(pdf_values)
# Condition covariance matrix a bit
# Should use attempt on multivariate_normal instead
min_eig = np.min(np.real(np.linalg.eigvals(covariance)))
while min_eig < 0:
print(Fore.YELLOW)
print("WARNING: Covariance matrix not positive-semidefinite")
print(f"Minimum eigenvalue: {min_eig}")
print("Introducing regulator...")
covariance -= 100 * min_eig * np.eye(*covariance.shape)
min_eig = np.min(np.real(np.linalg.eigvals(covariance)))
print(f"New minimum: {min_eig}")
return GPPDF(prior, mean, covariance, Q0, flavours, xs)
def MeasureNominal(t, mytree, pass_sys):
bpos_ = []
jetpos_ = []
jets_ = ROOT.vector('TLorentzVector')()
ordered_jets_ = ROOT.vector('TLorentzVector')()
bestcomb_ = []
corrcomb_ = []
corrb_ = []
corrlight_ = []
if (t.njets >= 7 and t.nBCSVM >= 2 and t.njets <= 14) or pass_sys:
if sample != 'data':
mytree.variables['weight'][0] = np.sign(t.genWeight)
mytree.variables['puweight'][0] = t.puWeight
mytree.variables['puweight_Up'][0] = t.puWeightUp
mytree.variables['puweight_Down'][0] = t.puWeightDown
mytree.variables['trigweight'][0] = t.triggerWeight
mytree.variables['trigweight_Up'][0] = t.triggerWeightUp
mytree.variables['trigweight_Down'][0] = t.triggerWeightDown
mytree.variables['btagweight'][0] = t.btagWeightCSV
mytree.variables['qgweight'][0] = t.qgWeight
mytree.variables['qgweight_Up'][0] = 2 * t.qgWeight - 1
mytree.variables['qgweight_Down'][0] = 1.0
mytree.variables['btagweight_jesPileUpPtBB_Down'][
0] = t.btagWeightCSV_down_jesPileUpPtBB
mytree.variables['btagweight_jesFlavorQCD_Down'][
0] = t.btagWeightCSV_down_jesFlavorQCD
mytree.variables['btagweight_jesAbsoluteScale_Down'][
0] = t.btagWeightCSV_down_jesAbsoluteScale
mytree.variables['btagweight_jesPileUpPtRef_Down'][
0] = t.btagWeightCSV_down_jesPileUpPtRef
mytree.variables['btagweight_jesRelativeFSR_Down'][
0] = t.btagWeightCSV_down_jesRelativeFSR
mytree.variables['btagweight_jesTimePtEta_Down'][
0] = t.btagWeightCSV_down_jesTimePtEta
mytree.variables['btagweight_hf_Down'][0] = t.btagWeightCSV_down_hf
mytree.variables['btagweight_cferr1_Down'][
0] = t.btagWeightCSV_down_cferr1
mytree.variables['btagweight_cferr2_Down'][
0] = t.btagWeightCSV_down_cferr2
mytree.variables['btagweight_jes_Down'][
0] = t.btagWeightCSV_down_jes
mytree.variables['btagweight_jesAbsoluteMPFBias_Down'][
0] = t.btagWeightCSV_down_jesAbsoluteMPFBias
mytree.variables['btagweight_lf_Down'][0] = t.btagWeightCSV_down_lf
mytree.variables['btagweight_jesPileUpPtEC1_Down'][
0] = t.btagWeightCSV_down_jesPileUpPtEC1
mytree.variables['btagweight_lfstats2_Down'][
0] = t.btagWeightCSV_down_lfstats2
mytree.variables['btagweight_lfstats1_Down'][
0] = t.btagWeightCSV_down_lfstats1
mytree.variables['btagweight_hfstats2_Down'][
0] = t.btagWeightCSV_down_hfstats2
mytree.variables['btagweight_hfstats1_Down'][
0] = t.btagWeightCSV_down_hfstats1
mytree.variables['btagweight_jesPileUpDataMC_Down'][
0] = t.btagWeightCSV_down_jesPileUpDataMC
mytree.variables['btagweight_jesPileUpPtBB_Up'][
0] = t.btagWeightCSV_up_jesPileUpPtBB
mytree.variables['btagweight_jesFlavorQCD_Up'][
0] = t.btagWeightCSV_up_jesFlavorQCD
mytree.variables['btagweight_jesAbsoluteScale_Up'][
0] = t.btagWeightCSV_up_jesAbsoluteScale
mytree.variables['btagweight_jesPileUpPtRef_Up'][
0] = t.btagWeightCSV_up_jesPileUpPtRef
mytree.variables['btagweight_jesRelativeFSR_Up'][
0] = t.btagWeightCSV_up_jesRelativeFSR
mytree.variables['btagweight_jesTimePtEta_Up'][
0] = t.btagWeightCSV_up_jesTimePtEta
mytree.variables['btagweight_hf_Up'][0] = t.btagWeightCSV_up_hf
mytree.variables['btagweight_cferr1_Up'][
0] = t.btagWeightCSV_up_cferr1
mytree.variables['btagweight_cferr2_Up'][
0] = t.btagWeightCSV_up_cferr2
mytree.variables['btagweight_jes_Up'][0] = t.btagWeightCSV_up_jes
mytree.variables['btagweight_jesAbsoluteMPFBias_Up'][
0] = t.btagWeightCSV_up_jesAbsoluteMPFBias
mytree.variables['btagweight_lf_Up'][0] = t.btagWeightCSV_up_lf
mytree.variables['btagweight_jesPileUpPtEC1_Up'][
0] = t.btagWeightCSV_up_jesPileUpPtEC1
mytree.variables['btagweight_hfstats2_Up'][
0] = t.btagWeightCSV_up_hfstats2
mytree.variables['btagweight_hfstats1_Up'][
0] = t.btagWeightCSV_up_hfstats1
mytree.variables['btagweight_lfstats2_Up'][
0] = t.btagWeightCSV_up_lfstats2
mytree.variables['btagweight_lfstats1_Up'][
0] = t.btagWeightCSV_up_lfstats1
mytree.variables['btagweight_jesPileUpDataMC_Up'][
0] = t.btagWeightCSV_up_jesPileUpDataMC
mytree.variables['LHEPDFweight'][0] = 1.0
mytree.variables['LHE_factweight'][0] = 1.0
mytree.variables['LHE_renormweight'][0] = 1.0
else:
mytree.variables['weight'][0] = 1
if 'ttbar' in sample:
nnpdfSet = lhapdf.getPDFSet("NNPDF30_nlo_as_0118")
pdfset = [1.0]
mytree.variables['LHE_factweight_Up'][0] = t.LHE_weights_scale_wgt[
0]
mytree.variables['LHE_factweight_Down'][
0] = t.LHE_weights_scale_wgt[1]
mytree.variables['LHE_renormweight_Up'][
0] = t.LHE_weights_scale_wgt[2]
mytree.variables['LHE_renormweight_Down'][
0] = t.LHE_weights_scale_wgt[3]
#for nlhe in range(t.nLHE_weights_scale):
# mytree.variables['LHE_scale'][nlhe]=t.LHE_weights_scale_wgt[nlhe]
for nlhe in range(t.nLHE_weights_pdf - 2):
pdfset.append(t.LHE_weights_pdf_wgt[nlhe])
pdfUnc = nnpdfSet.uncertainty(pdfset)
mytree.variables['LHEPDFweight_Up'][
0] = pdfUnc.central + pdfUnc.errplus
mytree.variables['LHEPDFweight_Down'][
0] = pdfUnc.central - pdfUnc.errminus
ntopjets = ntoptaggedjets = 0
nbjets = 0
nljets = 0
bdtcomb = -999999999
chi2comb = 999999999
mytree.variables['nPVs'][0] = t.nPVs
mytree.variables['sphericity'][0] = t.sphericity
mytree.variables['C'][0] = t.C
mytree.variables['minjetpt'][0] = t.jets_pt[t.njets - 1]
mytree.variables['jets_dRavg'][0] = sum(t.jets_dRave) / t.njets
drmin_list = [x for x in t.jets_dRmin if x > 0]
if len(drmin_list) > 0:
mytree.variables['jets_dRmin'][0] = min(
[x for x in t.jets_dRmin if x > 0])
else:
mytree.variables['jets_dRmin'][0] = 0
mytree.variables['jets_dRmax'][0] = max(t.jets_dRmax)
if sample == 'QCD': mytree.variables['isQCD'][0] = 1
else: mytree.variables['isQCD'][0] = 0
#mytree.variables['girth'][0] = ptR/(t.ht)**2
if 'ttbar' in sample or 'theory' in sample:
mytree.variables['ttCls'][0] = t.ttCls
else:
mytree.variables['ttCls'][0] = -1
mytree.variables['ht'][0] = t.ht
#mytree.variables['all_mass'][0] = t.invmass
#mytree.variables['closest_mass'][0] = t.mjjmin
mytree.variables['jet5pt'][0] = t.jets_pt[5]
mytree.variables['n_jets'][0] = t.njets
mytree.variables['n_bjets'][0] = t.nBCSVM
mytree.variables['n_addJets'][0] = t.njets - 6
#mytree.variables['n_topjets'][0] = ntopjets
mytree.variables['qgLR'][0] = -100
mytree.variables['centrality'][0] = t.centrality
mytree.variables['aplanarity'][0] = t.aplanarity
mytree.variables['meanDeltaRbtag'][0] = t.mean_dr_btag
mytree.variables['meanCSVbtag'][0] = t.mean_bdisc
mytree.variables['meanCSV'][0] = t.mean_bdisc_btag
mytree.variables['btagLR3b'][0] = t.btag_LR_3b_2b_btagCSV
mytree.variables['btagLR4b'][0] = t.btag_LR_4b_2b_btagCSV
mytree.variables['nBCSVM'][0] = t.nBCSVM
qgLR3b = t.qg_LR_3b_flavour_4q_0q
qgLR4b = t.qg_LR_4b_flavour_4q_0q
if t.nBCSVM == 3:
mytree.variables['qgLR'][0] = qgLR3b
# if t.mem_ttbb_FH_4w2h1t_p > 0: mytree.variables['memttbb'][0] = -TMath.Log10(t.mem_ttbb_FH_4w2h1t_p)
# else: mytree.variables['memttbb'][0] = -10
# #print mytree.variables['memttbb'][0]
# if nljets == 4:
# mytree.variables['qgLR'][0] = t.qg_LR_3b_flavour_4q_0q
# elif nljets >= 5:
# mytree.variables['qgLR'][0] = t.qg_LR_3b_flavour_5q_0q
if t.nBCSVM >= 4:
mytree.variables['qgLR'][0] = qgLR4b
# if t.mem_ttbb_FH_4w2h2t_p > 0: mytree.variables['memttbb'][0] = -TMath.Log10(t.mem_ttbb_FH_4w2h2t_p)
# else: mytree.variables['memttbb'][0] = -10
# #print mytree.variables['memttbb'][0]
# if nljets == 4: mytree.variables['qgLR'][0] = t.qg_LR_4b_flavour_4q_0q
# elif nljets >= 5: mytree.variables['qgLR'][0] = t.qg_LR_4b_flavour_5q_0q
else:
# mytree.variables['memttbb'][0] = -TMath.Log10(t.mem_ttbb_FH_4w2h1t_p) if t.mem_ttbb_FH_4w2h1t_p >0 else -10
mytree.variables['qgLR'][0] = qgLR3b
if sample != 'ttH':
tthcut = t.json and (t.HLT_ttH_FH or t.HLT_BIT_HLT_PFJet450_v
) and t.ht > 500 and t.jets_pt[5] > 40
else:
tthcut = t.json and (
t.HLT_ttH_FH) and t.ht > 500 and t.jets_pt[5] > 40
if sample == 'ttbar':
for itophad in range(t.ngenTopHad):
mytree.variables['genTopHad_pt'][itophad] = t.genTopHad_pt[
itophad]
mytree.variables['topweight'][0] = np.exp(
0.5 * (t.ngenTopLep * 0.0843616 - 0.000743051 *
np.sum(t.genTopLep_pt) + t.ngenTopHad * 0.0843616 -
0.000743051 * np.sum(t.genTopHad_pt))) * (t.jets_pt[0] /
t.jets_pt[0])
mytree.variables['topweight_Up'][0] = np.exp(0.5 * (
t.ngenTopLep * 0.00160296 - 0.000411375 * sum(t.genTopLep_pt) +
t.ngenTopHad * 0.00160296 - 0.000411375 * sum(t.genTopHad_pt)))
mytree.variables['topweight_Down'][0] = np.exp(
0.5 *
(t.ngenTopLep * 0.16712 - 0.00107473 * sum(t.genTopLep_pt) +
t.ngenTopHad * 0.16712 - 0.00107473 * sum(t.genTopHad_pt)))
if not (t.njets >= 7 and t.nBCSVM >= 2 and tthcut and t.njets <= 14):
return 0
for i in range(0, t.njets): #find the b-tagged jets
jet = ROOT.TLorentzVector()
jet.SetPtEtaPhiM(t.jets_pt[i], t.jets_eta[i], t.jets_phi[i],
t.jets_mass[i])
jets_.push_back(jet)
if sample != 'data':
if abs(t.jets_mcMatchId[i]) == 6:
if abs(t.jets_mcFlavour[i]
) == 5 and t.jets_matchFlag[i] == 1:
corrb_.append(i)
elif t.jets_matchFlag[i] == 0:
corrlight_.append(i)
alljets = range(t.njets)
if len(corrb_) == 2 and len(corrlight_) == 4:
mytree.variables['existCorrect'][0] = 1
corrcomb_ = JetComb(4, 2, corrlight_, corrb_)
for ii, c in enumerate(corrcomb_[0]):
mytree.variables['corr_comb'][ii] = c
else:
mytree.variables['existCorrect'][0] = 0
mytree.variables['hasbCorrect'][0] = 0
allcomb_ = JetComb(t.njets, 0, alljets)
mytree.variables['hasCorrect'][0] = 0
BDT_classes = [0, 0, 0]
printing = 0
highest = 0
bestcat = 0
#print comb_, 'comb_'
meanBDTttbar = sumprob = 0
maxbdt = 0
# print 'start comb'
# watch.Print()
for comb in allcomb_:
UpdateVariables(comb, jets_, mytree.variables, MVA_Only)
if mytree.variables['prob_chi2'][0] < probcut: continue
#if mytree.variables['simple_chi2'][0] < chi2cut: continue
if comb in corrcomb_:
mytree.variables['hasCorrect'][0] = 1
w1.UpdateParam(jets_[comb[2]] + jets_[comb[3]])
top1.UpdateParam(jets_[comb[0]] + jets_[comb[3]] + jets_[comb[2]])
w2.UpdateParam(jets_[comb[5]] + jets_[comb[4]])
top2.UpdateParam(jets_[comb[1]] + jets_[comb[4]] + jets_[comb[5]])
b1.UpdateParam(jets_[comb[0]])
b2.UpdateParam(jets_[comb[1]])
lp1.UpdateParam(jets_[comb[2]])
lp2.UpdateParam(jets_[comb[3]])
lq1.UpdateParam(jets_[comb[4]])
lq2.UpdateParam(jets_[comb[5]])
BDT_Value = reader.EvaluateMVA("BDT_Comb")
if BDT_Value > bdtcomb:
mytree.variables['BDT_Comb'][0] = BDT_Value
bdtcomb = BDT_Value
bestcomb_ = comb
if bestcomb_ in corrcomb_: mytree.variables['isCorrect'][0] = 1
else: mytree.variables['isCorrect'][0] = 0
# if sample !='data':
# mytree.variables['n_sumIDtop'][0] = t.jets_mcMatchId[comb[0]]
# mytree.variables['n_sumIDtop'][1] = t.jets_mcMatchId[comb[2]]
# mytree.variables['n_sumIDtop'][2] = t.jets_mcMatchId[comb[3]]
# mytree.variables['n_sumIDtop'][3] = t.jets_mcMatchId[comb[1]]
# mytree.variables['n_sumIDtop'][4] = t.jets_mcMatchId[comb[4]]
# mytree.variables['n_sumIDtop'][5] = t.jets_mcMatchId[comb[5]]
if len(bestcomb_) == 0: return 0
# print 'finish comb'
# watch.Print()
for ii, c in enumerate(bestcomb_):
mytree.variables['best_comb'][ii] = c
UpdateVariables(bestcomb_, jets_, mytree.variables, MVA_Only)
b1.UpdateParam(jets_[bestcomb_[0]])
b2.UpdateParam(jets_[bestcomb_[1]])
lp1.UpdateParam(jets_[bestcomb_[2]])
lp2.UpdateParam(jets_[bestcomb_[3]])
lq1.UpdateParam(jets_[bestcomb_[4]])
lq2.UpdateParam(jets_[bestcomb_[5]])
w1.UpdateParam(jets_[bestcomb_[2]] + jets_[bestcomb_[3]])
top1.UpdateParam(jets_[bestcomb_[0]] + jets_[bestcomb_[3]] +
jets_[bestcomb_[2]])
w2.UpdateParam(jets_[bestcomb_[5]] + jets_[bestcomb_[4]])
top2.UpdateParam(jets_[bestcomb_[1]] + jets_[bestcomb_[4]] +
jets_[bestcomb_[5]])
tt.UpdateParam(jets_[bestcomb_[0]] + jets_[bestcomb_[1]] +
jets_[bestcomb_[2]] + jets_[bestcomb_[3]] +
jets_[bestcomb_[4]] + jets_[bestcomb_[5]])
#print 'nominal'
#mytree.Print(useQCD)
mytree.variables['BDT_CWoLa'][0] = readerQCD.EvaluateMVA("BDT_QCD")
return 1
else:
return 0
def get_sys(input_h5_name):
input_h5 = h5py.File(input_h5_name, 'r')
print(list(input_h5.keys()))
print(list(input_h5['Histo1D'].keys()))
# print(input_h5[h1d_keyname+'/|MC_GENERIC|Pt'][:])
# for key in input_h5.keys():
# print(key)
observables = []
for key_name in all_data_keynames:
values = input_h5[key_name]
for value in values:
observables.append(value)
print("total {} observables".format(len(observables)))
variations = input_h5[variation_keyname]
n_variations = len(variations)
print("total {} variations".format(n_variations))
# print(variations[:])
# return
binids = input_h5[binid_keyname]
print(binids.shape)
print(binids[0])
n_bins = len(binids)
print("total {} bins".format(n_bins))
nominal_pdf_id = find_nomial_pdfid(input_h5)
nominal_pdf_name = pset_name(nominal_pdf_id)
scale_sys_name = "QCDScales"
n_exp_internal_pdf_vars = -1
pdfs_variations = collections.defaultdict(lambda: 0)
sys_variation_idx = collections.defaultdict(list)
for idx, variation in enumerate(variations):
if "MUR1_MUF1" in variation:
pdf_id = get_pdf_id(variation)
pdf_name = pset_name(pdf_id)
pdfs_variations[pdf_name] += 1
sys_variation_idx[pdf_name].append(idx)
if str(nominal_pdf_id) in variation:
sys_variation_idx[scale_sys_name].append(idx)
print("nominal PDF:", nominal_pdf_name)
n_external_pdfs = 0
for key, value in pdfs_variations.items():
print("{}: {} variations".format(key, value))
print("\t", sys_variation_idx[key])
if key != nominal_pdf_name:
n_external_pdfs += 1
print("{} external PDFs".format(n_external_pdfs))
install_pdfs(list(pdfs_variations.keys()))
pset = lhapdf.getPDFSet(nominal_pdf_name)
pdfs_ = pset.mkPDFs()
n_exp_internal_pdf_vars = len(pdfs_)
# evaluate the PDF sys
obs_values = get_obs_values(input_h5)
pdf_sys = np.full((n_bins, 3), fill_value=invalid_number, dtype=np.float32)
nominal_pdf = lhapdf.getPDFSet(nominal_pdf_name)
cl = 68
internal_pdf_sys = np.apply_along_axis(
lambda x: nominal_pdf.uncertainty(
x[sys_variation_idx[nominal_pdf_name]], cl).errsymm, 1, obs_values)
print(internal_pdf_sys[1])
print(internal_pdf_sys.shape)
# external PDF sets
external_pdf_values = np.zeros((n_bins, n_external_pdfs + 1),
dtype=np.float32)
inorm = -1
idx = 0
for key, value in pdfs_variations.items():
print(key, value)
if key == nominal_pdf_name:
inorm = idx
if value > 1:
this_pdf = lhapdf.getPDFSet(key)
external_pdf_values[:, idx] = np.apply_along_axis(
lambda x: this_pdf.uncertainty(x[sys_variation_idx[key]], cl).
central, 1, obs_values)
else:
external_pdf_values[:, idx] = obs_values[:,
sys_variation_idx[key][0]]
idx += 1
external_diff = external_pdf_values - external_pdf_values[:,
inorm].reshape(
-1, 1)
# print(np.any(np.nonzero(external_diff[:, inorm])))
external_pdf_sys = np.amax(external_diff, axis=1)
pdf_sys[:, 0] = internal_pdf_sys
pdf_sys[:, 1] = external_pdf_sys
pdf_sys[:, 2] = np.amax(pdf_sys[:, 0:2], axis=1)
scale_diff = obs_values[:, sys_variation_idx[
scale_sys_name]] - obs_values[:, 0].reshape(-1, 1)
scale_sys = np.amax(scale_diff, axis=1)
out_file = h5py.File("Rivet_pdfsys.h5", 'w')
out_file.create_dataset("pdfsys", data=pdf_sys, dtype=np.float32)
out_file.create_dataset("scalesys", data=scale_sys, dtype=np.float32)
out_file.close()
input_h5.close()
def lhapdf_xfs(pdf_name, xs, flavors, Q2):
pdf_information = lhapdf.getPDFSet(pdf_name)
n_xs = len(xs)
Q = np.sqrt(Q2)
if (pdf_information.errorType == 'hessian') or (pdf_information.errorType
== 'symmhessian'):
## 'lhapdf.getPDFSet("ABMP16_3_nlo").errorType' is 'symmhessian'
pdf_0 = lhapdf.mkPDF(pdf_name, 0)
xfs = {}
for flavor in flavors:
xfs[flavor] = {
'center': np.zeros(n_xs),
'difference': np.zeros(n_xs)
}
for i in range(1, ((pdf_information.size - 1) / 2) + 1):
pdf_plus = lhapdf.mkPDF(pdf_name, 2 * i)
pdf_minus = lhapdf.mkPDF(pdf_name, 2 * i - 1)
for j in range(n_xs):
for flavor in flavors:
if flavor == 'uv':
xfs[flavor]['center'][j] = pdf_0.xfxQ(
2, xs[j], Q) - pdf_0.xfxQ(-2, xs[j], Q)
xfs[flavor]['difference'][j] += (
pdf_plus.xfxQ(2, xs[j], Q) -
pdf_plus.xfxQ(-2, xs[j], Q) -
(pdf_minus.xfxQ(2, xs[j], Q) -
pdf_minus.xfxQ(-2, xs[j], Q)))**2.0
elif flavor == 'dv':
xfs[flavor]['center'][j] = pdf_0.xfxQ(
1, xs[j], Q) - pdf_0.xfxQ(-1, xs[j], Q)
xfs[flavor]['difference'][j] += (
pdf_plus.xfxQ(1, xs[j], Q) -
pdf_plus.xfxQ(-1, xs[j], Q) -
(pdf_minus.xfxQ(1, xs[j], Q) -
pdf_minus.xfxQ(-1, xs[j], Q)))**2.0
elif flavor == 'u':
xfs[flavor]['center'][j] = pdf_0.xfxQ(2, xs[j], Q)
xfs[flavor]['difference'][j] += (
pdf_plus.xfxQ(2, xs[j], Q) -
pdf_minus.xfxQ(2, xs[j], Q))**2.0
elif flavor == 'd':
xfs[flavor]['center'][j] = pdf_0.xfxQ(1, xs[j], Q)
xfs[flavor]['difference'][j] += (
pdf_plus.xfxQ(1, xs[j], Q) -
pdf_minus.xfxQ(1, xs[j], Q))**2.0
elif flavor == 'd/u':
xfs[flavor]['center'][j] = pdf_0.xfxQ(
1, xs[j], Q) / pdf_0.xfxQ(2, xs[j], Q)
xfs[flavor]['difference'][j] += (
(pdf_plus.xfxQ(1, xs[j], Q) -
pdf_minus.xfxQ(1, xs[j], Q)) /
pdf_0.xfxQ(2, xs[j], Q))**2.0
elif flavor == 'db+ub':
xfs[flavor]['center'][j] = pdf_0.xfxQ(
-1, xs[j], Q) + pdf_0.xfxQ(-2, xs[j], Q)
xfs[flavor]['difference'][j] += (
pdf_plus.xfxQ(-1, xs[j], Q) +
pdf_plus.xfxQ(-2, xs[j], Q) -
(pdf_minus.xfxQ(-1, xs[j], Q) +
pdf_minus.xfxQ(-2, xs[j], Q)))**2.0
elif flavor == 'db-ub':
xfs[flavor]['center'][j] = pdf_0.xfxQ(
-1, xs[j], Q) - pdf_0.xfxQ(-2, xs[j], Q)
xfs[flavor]['difference'][j] += (
pdf_plus.xfxQ(-1, xs[j], Q) -
pdf_plus.xfxQ(-2, xs[j], Q) -
(pdf_minus.xfxQ(-1, xs[j], Q) -
pdf_minus.xfxQ(-2, xs[j], Q)))**2.0
elif flavor == 's+sb':
xfs[flavor]['center'][j] = pdf_0.xfxQ(
3, xs[j], Q) + pdf_0.xfxQ(-3, xs[j], Q)
xfs[flavor]['difference'][j] += (
pdf_plus.xfxQ(3, xs[j], Q) +
pdf_plus.xfxQ(-3, xs[j], Q) -
(pdf_minus.xfxQ(3, xs[j], Q) +
pdf_minus.xfxQ(-3, xs[j], Q)))**2.0
elif flavor == 'g':
xfs[flavor]['center'][j] = pdf_0.xfxQ(21, xs[j], Q)
xfs[flavor]['difference'][j] += (
pdf_plus.xfxQ(21, xs[j], Q) -
pdf_minus.xfxQ(21, xs[j], Q))**2.0
elif flavor == 'rs':
if pdf_name == 'CJ15nlo':
## R_s is set to be 0.4 in CJ15
xfs[flavor]['center'][j] = 0.4
xfs[flavor]['difference'][j] += 0.0
else:
xfs[flavor]['center'][j] = (pdf_0.xfxQ(
3, xs[j], Q) + pdf_0.xfxQ(-3, xs[j], Q)) / (
pdf_0.xfxQ(-1, xs[j], Q) +
pdf_0.xfxQ(-2, xs[j], Q))
xfs[flavor]['difference'][j] += ((pdf_plus.xfxQ(3, xs[j], Q) + pdf_plus.xfxQ(-3, xs[j], Q) - (pdf_minus.xfxQ(3, xs[j], Q) + pdf_minus.xfxQ(-3, xs[j], Q))) / \
(pdf_0.xfxQ(-1, xs[j], Q) + pdf_0.xfxQ(-2, xs[j], Q))) ** 2.0
for flavor in flavors:
xfs[flavor]['difference'] = 0.5 * np.sqrt(
xfs[flavor]['difference'])
elif pdf_information.errorType == 'replicas':
pdf_all = lhapdf.mkPDFs(pdf_name)
xfs = {}
xfs_temp = {}
for flavor in flavors:
xfs[flavor] = {'center': [], 'difference': []}
for i in range(n_xs):
for flavor in flavors:
xfs_temp[flavor] = []
for pdf_individual in pdf_all:
for flavor in flavors:
if flavor == 'uv':
xfs_temp[flavor].append(
pdf_individual.xfxQ(2, xs[i], Q) -
pdf_individual.xfxQ(-2, xs[i], Q))
elif flavor == 'dv':
xfs_temp[flavor].append(
pdf_individual.xfxQ(1, xs[i], Q) -
pdf_individual.xfxQ(-1, xs[i], Q))
elif flavor == 'u':
xfs_temp[flavor].append(
pdf_individual.xfxQ(2, xs[i], Q))
elif flavor == 'd':
xfs_temp[flavor].append(
pdf_individual.xfxQ(1, xs[i], Q))
elif flavor == 'd/u':
xfs_temp[flavor].append(
pdf_individual.xfxQ(1, xs[i], Q) /
pdf_individual.xfxQ(2, xs[i], Q))
elif flavor == 'db+ub':
xfs_temp[flavor].append(
pdf_individual.xfxQ(-1, xs[i], Q) +
pdf_individual.xfxQ(-2, xs[i], Q))
elif flavor == 'db-ub':
xfs_temp[flavor].append(
pdf_individual.xfxQ(-1, xs[i], Q) -
pdf_individual.xfxQ(-2, xs[i], Q))
elif flavor == 's+sb':
xfs_temp[flavor].append(
pdf_individual.xfxQ(3, xs[i], Q) +
pdf_individual.xfxQ(-3, xs[i], Q))
elif flavor == 'g':
xfs_temp[flavor].append(
pdf_individual.xfxQ(21, xs[i], Q))
elif flavor == 'rs':
xfs_temp[flavor].append(
(pdf_individual.xfxQ(3, xs[i], Q) +
pdf_individual.xfxQ(-3, xs[i], Q)) /
(pdf_individual.xfxQ(-1, xs[i], Q) +
pdf_individual.xfxQ(-2, xs[i], Q)))
for flavor in flavors:
xfs[flavor]['center'].append(np.mean(xfs_temp[flavor], axis=0))
xfs[flavor]['difference'].append(
np.std(xfs_temp[flavor], axis=0))
for flavor in flavors:
xfs[flavor]['center'] = np.array(xfs[flavor]['center'])
xfs[flavor]['difference'] = np.array(xfs[flavor]['difference'])
return xfs
def __missing__(self, key):
el = self[key] = lhapdf.getPDFSet(key)
return el
#! /usr/bin/env python
## Python LHAPDF6 usage example for PDF uncertainty code (G. Watt, 24/04/2014).
## Extended in July 2015 for ErrorType values ending in "+as".
## Modified in September 2015 for more general ErrorType values:
## number of parameter variations determined by counting "+" symbols.
from __future__ import print_function, division
import lhapdf
x = 0.1
q = 100.0
pset = lhapdf.getPDFSet("CT10nnlo") # ErrorType: hessian
#pset = lhapdf.getPDFSet("abm12lhc_5_nnlo") # ErrorType: symmhessian
#pset = lhapdf.getPDFSet("NNPDF30_nnlo_as_0118") # ErrorType: replicas
#pset = lhapdf.getPDFSet("NNPDF30_nnlo_nf_5_pdfas") # ErrorType: replicas+as
pdfs = pset.mkPDFs()
nmem = pset.size - 1
print()
print("Error type = ", pset.errorType)
print("Error conf level = ", pset.errorConfLevel)
print()
npar = pset.errorType.count(
"+") # number of parameter variations (alphaS, etc.)
if npar > 0:
print("Last %d members are parameter variations\n" % (2 * npar))
## Fill vectors xgAll and xuAll using all PDF members.
xgAll = [0.0 for i in range(pset.size)]
"CITo2Mu_Lam40TeVConLL_M800to1300" : 0.014510,
"CITo2Mu_Lam40TeVConLR_M800to1300" : 0.014520,
"CITo2Mu_Lam40TeVConRR_M800to1300" : 0.014540,
"CITo2Mu_Lam40TeVDesLL_M800to1300" : 0.013790,
"CITo2Mu_Lam40TeVDesLR_M800to1300" : 0.013730,
"CITo2Mu_Lam40TeVDesRR_M800to1300" : 0.013750,
'2016_CITo2Mu_Lam16TeVDesRR_M300':0.550000,
'2016_CITo2Mu_Lam16TeVDesRR_M800':0.013670,
'2016_CITo2Mu_Lam16TeVDesRR_M1300':0.002403,
'2016_CITo2Mu_Lam16TeVConRR_M300':0.583500,
'2016_CITo2Mu_Lam16TeVConRR_M800':0.019120,
'2016_CITo2Mu_Lam16TeVConRR_M1300':0.003939,
}
pdfsetRef = lhapdf.getPDFSet("NNPDF31_nnlo_as_0118")
pdfset31LO = lhapdf.getPDFSet("NNPDF31_lo_as_0118")
pdfset31LO13 = lhapdf.getPDFSet("NNPDF31_lo_as_0130")
pdfset30NNLO = lhapdf.getPDFSet("NNPDF30_nnlo_as_0118")
pdfset30LO = lhapdf.getPDFSet("NNPDF30_lo_as_0118")
pdfset30LO13 = lhapdf.getPDFSet("NNPDF30_lo_as_0130")
pdfset23LO = lhapdf.getPDFSet("NNPDF23_lo_as_0130_qed")
# ~ print (pdfset.description)
pdfReplicasRef = pdfsetRef.mkPDFs()
pdfReplicas31LO = pdfset31LO.mkPDFs()
pdfReplicas31LO13 = pdfset31LO13.mkPDFs()
pdfReplicas30LO = pdfset30LO.mkPDFs()
pdfReplicas30LO13 = pdfset30LO13.mkPDFs()
pdfReplicas30NNLO = pdfset30NNLO.mkPDFs()
pdfReplicas23LO = pdfset23LO.mkPDFs()
w = 0.00407 #Higgs total decay width
# w = 0.00374 #Higgs total decay width
Ncg = 8 #Number of colors for gluon
Nsg = 2 #Number of spin projections for gluon
E = sqrt(shat)
bw = 1./Ncg**2 * 1./Nsg**2 * 16*pi/shat * w**2/4/( (E-mH)**2 + w**2/4) * Bin * Bout
bw *= 2 #magic factor due to identical gluons
return bw * ge2mb
# Define PDF to be used
name = "MRST2007lomod"
name = "MRST2004qed_proton"
pdf = lhapdf.getPDFSet(name).mkPDF(0)
# Define histograms to be filed, `dmH` half of the generated mass widht, `dy` is half of the rapidity interval of the interest
dmH = 0.100
dy = 5
hEta = TH1D("eta", "Eta of H", 20, -dy, dy)
hMass = TH1D("mass", "Mass of H", 61, mH-dmH, mH+dmH)
# Define the beam energy of the protons
Eb = 6500
# Define the number of points to be generated
npoints = 1200000
#! /usr/bin/env python
## Python LHAPDF6 usage example for PDF uncertainty code (G. Watt, 24/04/2014)
import lhapdf
x = 0.1
q = 100.0
pset = lhapdf.getPDFSet("CT10nnlo")
#pset = lhapdf.getPDFSet("NNPDF23_nnlo_as_0118")
pdfs = pset.mkPDFs()
nmem = pset.size - 1
print
print "Error type = ", pset.errorType
print "Error conf level = ", pset.errorConfLevel
## Fill vectors xgAll and xuAll using all PDF members.
xgAll = [0.0 for i in xrange(pset.size)]
xuAll = [0.0 for i in xrange(pset.size)]
for imem in xrange(pset.size):
xgAll[imem] = pdfs[imem].xfxQ(21, x, q)
xuAll[imem] = pdfs[imem].xfxQ(2, x, q)
# print imem, xgAll[imem], xuAll[imem]
## Calculate 1-sigma PDF uncertainty on gluon distribution.
unc = pset.uncertainty(xgAll)
print "xg = %.4g + %.4g - %.4g (+- %.4g)" % (unc.central, unc.errplus, unc.errminus, unc.errsymm)
print "Scale = %.4g" % unc.scale
## Calculate 1-sigma PDF uncertainty on up-quark distribution.
@author: hkaveh
"""
import nuSolutions as nu
import ROOT as r
import lhapdf
import numpy as np
import math
from operator import itemgetter
print lhapdf.version()
lhapdf.pathsPrepend("/home/hkaveh/")
print lhapdf.availablePDFSets()
print lhapdf.paths()
pset = lhapdf.getPDFSet('NNPDF23_nlo_as_0118')
print pset.description
p = lhapdf.mkPDF('NNPDF23_nlo_as_0118')
def get_dalitz_prob(lep,top,mb,mw):
mte = lep.Dot( top );
mt = top.M();
mt2 = mt * mt;
mb2 = mb * mb;
mw2 = mw * mw;
mt2_mb2 = mt2 - mb2;
return 4. * mt*lep.E() * ( mt2 - mb2 - 2. * mt*lep.E() ) /( mt2_mb2 * mt2_mb2 + mw2 * ( mt2 - mb2 ) - 2. * mw2 * mw2 )
def getweight(lep_p,lep_m,nu1,nu2,bquark1,bquark2):
t1 = lep_p + nu1 + bquark1;
def MeasureNominal(t, mytree, pass_sys):
bpos_ = []
jetpos_ = []
jets_ = ROOT.vector('TLorentzVector')()
ordered_jets_ = ROOT.vector('TLorentzVector')()
bestcomb_ = []
corrcomb_ = []
corrb_ = []
corrlight_ = []
if (t.njets >= 7 and t.nBCSVM >= 2) or pass_sys:
if sample != 'data':
mytree.variables['weight'][0] = np.sign(t.genWeight)
mytree.variables['puweight'][0] = t.puWeight
mytree.variables['puweight_Up'][0] = t.puWeightUp
mytree.variables['puweight_Down'][0] = t.puWeightDown
mytree.variables['trigweight'][0] = t.triggerWeight
mytree.variables['trigweight_Up'][0] = t.triggerWeightUp
mytree.variables['trigweight_Down'][0] = t.triggerWeightDown
mytree.variables['btagweight'][0] = t.btagWeightCSV
mytree.variables['qgweight'][0] = t.qgWeight
mytree.variables['qgweight_Up'][0] = 2 * t.qgWeight - 1
mytree.variables['qgweight_Down'][0] = 1.0
mytree.variables['btagweight_jesPileUpPtBB_Down'][
0] = t.btagWeightCSV_down_jesPileUpPtBB
mytree.variables['btagweight_jesFlavorQCD_Down'][
0] = t.btagWeightCSV_down_jesFlavorQCD
mytree.variables['btagweight_jesAbsoluteScale_Down'][
0] = t.btagWeightCSV_down_jesAbsoluteScale
mytree.variables['btagweight_jesPileUpPtRef_Down'][
0] = t.btagWeightCSV_down_jesPileUpPtRef
mytree.variables['btagweight_jesRelativeFSR_Down'][
0] = t.btagWeightCSV_down_jesRelativeFSR
mytree.variables['btagweight_jesTimePtEta_Down'][
0] = t.btagWeightCSV_down_jesTimePtEta
mytree.variables['btagweight_hf_Down'][0] = t.btagWeightCSV_down_hf
mytree.variables['btagweight_cferr1_Down'][
0] = t.btagWeightCSV_down_cferr1
mytree.variables['btagweight_cferr2_Down'][
0] = t.btagWeightCSV_down_cferr2
mytree.variables['btagweight_jes_Down'][
0] = t.btagWeightCSV_down_jes
mytree.variables['btagweight_jesAbsoluteMPFBias_Down'][
0] = t.btagWeightCSV_down_jesAbsoluteMPFBias
mytree.variables['btagweight_lf_Down'][0] = t.btagWeightCSV_down_lf
mytree.variables['btagweight_jesPileUpPtEC1_Down'][
0] = t.btagWeightCSV_down_jesPileUpPtEC1
mytree.variables['btagweight_lfstats2_Down'][
0] = t.btagWeightCSV_down_lfstats2
mytree.variables['btagweight_lfstats1_Down'][
0] = t.btagWeightCSV_down_lfstats1
mytree.variables['btagweight_hfstats2_Down'][
0] = t.btagWeightCSV_down_hfstats2
mytree.variables['btagweight_hfstats1_Down'][
0] = t.btagWeightCSV_down_hfstats1
mytree.variables['btagweight_jesPileUpDataMC_Down'][
0] = t.btagWeightCSV_down_jesPileUpDataMC
mytree.variables['btagweight_jesPileUpPtBB_Up'][
0] = t.btagWeightCSV_up_jesPileUpPtBB
mytree.variables['btagweight_jesFlavorQCD_Up'][
0] = t.btagWeightCSV_up_jesFlavorQCD
mytree.variables['btagweight_jesAbsoluteScale_Up'][
0] = t.btagWeightCSV_up_jesAbsoluteScale
mytree.variables['btagweight_jesPileUpPtRef_Up'][
0] = t.btagWeightCSV_up_jesPileUpPtRef
mytree.variables['btagweight_jesRelativeFSR_Up'][
0] = t.btagWeightCSV_up_jesRelativeFSR
mytree.variables['btagweight_jesTimePtEta_Up'][
0] = t.btagWeightCSV_up_jesTimePtEta
mytree.variables['btagweight_hf_Up'][0] = t.btagWeightCSV_up_hf
mytree.variables['btagweight_cferr1_Up'][
0] = t.btagWeightCSV_up_cferr1
mytree.variables['btagweight_cferr2_Up'][
0] = t.btagWeightCSV_up_cferr2
mytree.variables['btagweight_jes_Up'][0] = t.btagWeightCSV_up_jes
mytree.variables['btagweight_jesAbsoluteMPFBias_Up'][
0] = t.btagWeightCSV_up_jesAbsoluteMPFBias
mytree.variables['btagweight_lf_Up'][0] = t.btagWeightCSV_up_lf
mytree.variables['btagweight_jesPileUpPtEC1_Up'][
0] = t.btagWeightCSV_up_jesPileUpPtEC1
mytree.variables['btagweight_lfstats2_Up'][
0] = t.btagWeightCSV_up_lfstats2
mytree.variables['btagweight_lfstats1_Up'][
0] = t.btagWeightCSV_up_lfstats1
mytree.variables['btagweight_hfstats2_Up'][
0] = t.btagWeightCSV_up_hfstats2
mytree.variables['btagweight_hfstats1_Up'][
0] = t.btagWeightCSV_up_hfstats1
mytree.variables['btagweight_jesPileUpDataMC_Up'][
0] = t.btagWeightCSV_up_jesPileUpDataMC
mytree.variables['LHEPDFweight'][0] = 1.0
mytree.variables['LHE_factweight'][0] = 1.0
mytree.variables['LHE_renormweight'][0] = 1.0
else:
mytree.variables['weight'][0] = 1
ntopjets = ntoptaggedjets = 0
if sample == 'ttbar':
nnpdfSet = lhapdf.getPDFSet("NNPDF30_nlo_as_0118")
pdfset = [1.0]
mytree.variables['LHE_factweight_Up'][0] = t.LHE_weights_scale_wgt[
0]
mytree.variables['LHE_factweight_Down'][
0] = t.LHE_weights_scale_wgt[1]
mytree.variables['LHE_renormweight_Up'][
0] = t.LHE_weights_scale_wgt[2]
mytree.variables['LHE_renormweight_Down'][
0] = t.LHE_weights_scale_wgt[3]
#for nlhe in range(t.nLHE_weights_scale):
# mytree.variables['LHE_scale'][nlhe]=t.LHE_weights_scale_wgt[nlhe]
for nlhe in range(t.nLHE_weights_pdf - 2):
pdfset.append(t.LHE_weights_pdf_wgt[nlhe])
pdfUnc = nnpdfSet.uncertainty(pdfset)
mytree.variables['LHEPDFweight_Up'][
0] = pdfUnc.central + pdfUnc.errplus
mytree.variables['LHEPDFweight_Down'][
0] = pdfUnc.central - pdfUnc.errminus
mytree.variables['nPVs'][0] = t.nPVs
mytree.variables['sphericity'][0] = t.sphericity
mytree.variables['C'][0] = t.C
mytree.variables['minjetpt'][0] = t.jets_pt[t.njets - 1]
mytree.variables['jets_dRavg'][0] = sum(t.jets_dRave) / t.njets
drmin_list = [x for x in t.jets_dRmin if x > 0]
if len(drmin_list) > 0:
mytree.variables['jets_dRmin'][0] = min(
[x for x in t.jets_dRmin if x > 0])
else:
mytree.variables['jets_dRmin'][0] = 0
mytree.variables['jets_dRmax'][0] = max(t.jets_dRmax)
if sample == 'QCD': mytree.variables['isQCD'][0] = 1
else: mytree.variables['isQCD'][0] = 0
#mytree.variables['girth'][0] = ptR/(t.ht)**2
if 'ttbar' in sample or 'theory' in sample:
mytree.variables['ttCls'][0] = t.ttCls
else:
mytree.variables['ttCls'][0] = -1
mytree.variables['ht'][0] = t.ht
#mytree.variables['all_mass'][0] = t.invmass
#mytree.variables['closest_mass'][0] = t.mjjmin
mytree.variables['jet5pt'][0] = t.jets_pt[5]
mytree.variables['n_jets'][0] = t.njets
mytree.variables['n_bjets'][0] = t.nBCSVM
mytree.variables['n_addJets'][0] = t.njets - 6
#mytree.variables['n_topjets'][0] = ntopjets
mytree.variables['qgLR'][0] = -100
mytree.variables['centrality'][0] = t.centrality
mytree.variables['aplanarity'][0] = t.aplanarity
mytree.variables['meanDeltaRbtag'][0] = t.mean_dr_btag
mytree.variables['meanCSVbtag'][0] = t.mean_bdisc
mytree.variables['meanCSV'][0] = t.mean_bdisc_btag
mytree.variables['btagLR3b'][0] = t.btag_LR_3b_2b_btagCSV
mytree.variables['btagLR4b'][0] = t.btag_LR_4b_2b_btagCSV
mytree.variables['nBCSVM'][0] = t.nBCSVM
mytree.variables['chi2'][0] = t.chi2
mytree.variables['prob_chi2'][0] = TMath.Prob(t.chi2, 4)
qgLR3b = t.qg_LR_3b_flavour_4q_0q
qgLR4b = t.qg_LR_4b_flavour_4q_0q
if t.nBCSVM == 3:
mytree.variables['qgLR'][0] = qgLR3b
# if t.mem_ttbb_FH_4w2h1t_p > 0: mytree.variables['memttbb'][0] = -TMath.Log10(t.mem_ttbb_FH_4w2h1t_p)
# else: mytree.variables['memttbb'][0] = -10
# #print mytree.variables['memttbb'][0]
# if nljets == 4:
# mytree.variables['qgLR'][0] = t.qg_LR_3b_flavour_4q_0q
# elif nljets >= 5:
# mytree.variables['qgLR'][0] = t.qg_LR_3b_flavour_5q_0q
if t.nBCSVM >= 4:
mytree.variables['qgLR'][0] = qgLR4b
# if t.mem_ttbb_FH_4w2h2t_p > 0: mytree.variables['memttbb'][0] = -TMath.Log10(t.mem_ttbb_FH_4w2h2t_p)
# else: mytree.variables['memttbb'][0] = -10
# #print mytree.variables['memttbb'][0]
# if nljets == 4: mytree.variables['qgLR'][0] = t.qg_LR_4b_flavour_4q_0q
# elif nljets >= 5: mytree.variables['qgLR'][0] = t.qg_LR_4b_flavour_5q_0q
else:
# mytree.variables['memttbb'][0] = -TMath.Log10(t.mem_ttbb_FH_4w2h1t_p) if t.mem_ttbb_FH_4w2h1t_p >0 else -10
mytree.variables['qgLR'][0] = qgLR3b
if sample != 'ttH':
tthcut = t.json and (t.HLT_ttH_FH or t.HLT_BIT_HLT_PFJet450_v
) and t.ht > 500 and t.jets_pt[5] > 40
else:
tthcut = t.json and (
t.HLT_ttH_FH) and t.ht > 500 and t.jets_pt[5] > 40
mytree.variables['hasCorrect'][0] = -1
printing = 0
highest = 0
bestcat = 0
#print comb_, 'comb_'
meanBDTttbar = sumprob = 0
maxbdt = 0
# print 'start comb'
# watch.Print()
if 'ttbar' in sample:
for itophad in range(t.ngenTopHad):
mytree.variables['genTopHad_pt'][itophad] = t.genTopHad_pt[
itophad]
mytree.variables['topweight'][0] = np.exp(
0.5 * (t.ngenTopLep * 0.0843616 - 0.000743051 *
np.sum(t.genTopLep_pt) + t.ngenTopHad * 0.0843616 -
0.000743051 * np.sum(t.genTopHad_pt))) * (t.jets_pt[0] /
t.jets_pt[0])
mytree.variables['topweight_Up'][0] = np.exp(0.5 * (
t.ngenTopLep * 0.00160296 - 0.000411375 * sum(t.genTopLep_pt) +
t.ngenTopHad * 0.00160296 - 0.000411375 * sum(t.genTopHad_pt)))
mytree.variables['topweight_Down'][0] = np.exp(
0.5 *
(t.ngenTopLep * 0.16712 - 0.00107473 * sum(t.genTopLep_pt) +
t.ngenTopHad * 0.16712 - 0.00107473 * sum(t.genTopHad_pt)))
if not (t.njets >= 7 and t.nBCSVM >= 2 and tthcut): return 0
mytree.variables['BDT_CWoLa'][0] = t.CWoLa_BDT
mytree.variables['BDT_Comb'][0] = t.perm_BDT
return 1
else:
return 0
import lhapdf
## Getting a PDF member object
p = lhapdf.mkPDF("CT10", 0)
p = lhapdf.mkPDF("CT10/0")
## Gluon PDF querying at x=0.001, Q2=10000
print(p.xfxQ2(21, 1e-3, 1e4))
## Basic all-flavour PDF querying at x=0.01, Q=M_Z
for pid in p.flavors():
print(p.xfxQ(pid, 0.01, 91.2))
# TODO: demonstrate looping over PDF set members
pset = lhapdf.getPDFSet("CT10nlo")
print(pset.description)
pcentral = pset.mkPDF(0)
pdfs1 = pset.mkPDFs()
pdfs2 = lhapdf.mkPDFs("CT10nlo") # a direct way to get all the set's PDFs
## Filling a numpy 2D array with xf(x,Q) samples
import numpy as np
xs = [x for x in np.logspace(-7, 0, 5)]
qs = [q for q in np.logspace(1, 4, 4)]
gluon_xfs = np.empty([len(xs), len(qs)])
for ix, x in enumerate(xs):
for iq, q in enumerate(qs):
gluon_xfs[ix, iq] = p.xfxQ(21, x, q)
print(gluon_xfs)
