def nevents_plot():
in_f = ROOT.TFile(gp.fn)
out_f = ROOT.TFile('nevents.root', 'recreate')
kinds, masses, taus, nmasses, ntaus = axes(in_f)
for kind in kinds:
h = ROOT.TH2D('nevents_%s' % kind, ';mass (GeV);#tau (mm)', nmasses, masses, ntaus, taus)
for ibin in xrange(1, nmasses+1):
mass = h.GetXaxis().GetBinLowEdge(ibin)
for jbin in xrange(1, ntaus+1):
tau = h.GetYaxis().GetBinLowEdge(jbin)
try:
isample = name2isample(in_f, details2name(kind, tau, mass))
except ValueError:
continue
nev = nevents(in_f, isample)
h.SetBinContent(ibin, jbin, nev)
h.SetBinError (ibin, jbin, nev**0.5)
out_f.cd()
h.Write()
out_f.Write()
out_f.Close()
def event_scale_factors():
from signal_efficiency import SignalEfficiencyCombiner
for which, years in ('2017p8', [2017,2018]), :
combiner = SignalEfficiencyCombiner(years)
in_f = combiner.inputs[0].f
out_f = ROOT.TFile('event_TrackMover_scale_factors_%s.root' % which, 'recreate')
kinds, masses, taus, nmasses, ntaus = axes(in_f)
taus.remove(30.)
ntaus -= 1
for kind in kinds:
h = ROOT.TH2D('event_TrackMover_scale_factors_%s' % kind, ';mass (GeV);#tau (mm)', nmasses, masses, ntaus, taus)
for ibin in xrange(1, nmasses+1):
mass = h.GetXaxis().GetBinLowEdge(ibin)
for jbin in xrange(1, ntaus+1):
tau = h.GetYaxis().GetBinLowEdge(jbin)
pt = kind, tau, mass
try:
name = details2name(*pt)
isample = name2isample(in_f, name)
except ValueError as exc:
if tau < 100:
print colors.warning('problem getting %r : isample %r exc %s' % (pt, isample, exc))
continue # leave holes in the plot to know which samples are missing
event_SF = None
if which == '2017p8' :
vtx_SF_2017 = sig_datamcSF_2017p8(name+"_2017")
vtx_SF_2018 = sig_datamcSF_2017p8(name+"_2018")
event_SF_2017 = vtx_SF_2017**2
event_SF_2018 = vtx_SF_2018**2
lumi2017 = gp.int_lumis[gp.years.index("2017")]
lumi2018 = gp.int_lumis[gp.years.index("2018")]
event_SF = (event_SF_2017*lumi2017 + event_SF_2018*lumi2018) / (lumi2017+lumi2018)
# for table of SFs, with bins 0.1--0.3\mm & 0.3--1\mm & 1--10\mm & 10--100\mm,
# take the ~midpoint of each bin and put its SF in the table (averaged where necessary)
if mass == 1600 :
if tau == 0.2 or tau == 0.6 or tau == 0.7 or tau == 4 or tau == 7 or tau == 55 :
print("for SF table:",name, round(event_SF, 4))
else :
sys.exit('event_scale_factors not set up to handle %s! Exiting.' % which)
h.SetBinContent(ibin, jbin, event_SF)
h.SetBinError (ibin, jbin, 0) # JMTBAD
out_f.cd()
h.Write()
out_f.Write()
out_f.Close()
def make():
def warning():
return
for i in xrange(20):
print colors.error("don't forget: anything?")
warning()
assert not os.path.exists(gp.fn)
ROOT.TH1.AddDirectory(0)
f = ROOT.TFile(gp.fn, 'recreate')
make_bkg(f)
name_list = {}
make_signals_2015p6(f, name_list)
make_signals_2017p8(f, name_list)
title = name_list.pop('title')
isamples = sorted([s.isample for s in name_list.itervalues()],
reverse=True)
assert isamples == range(-1, isamples[-1] - 1, -1)
nsigs = len(isamples)
h = ROOT.TH1C('name_list', title, nsigs, 0, nsigs)
missing = defaultdict(list)
for s in name_list.itervalues():
miss = tuple(sorted(set(gp.years) - s.years))
missing[miss].append(s.name)
c = sum((int(year in s.years) << y) for y, year in enumerate(gp.years))
h.GetXaxis().SetBinLabel(-s.isample, s.name)
h.SetBinContent(-s.isample, c)
f.cd()
h.Write()
f.Close()
if missing:
w = colors.warning
print w('missing summary:')
for k in sorted(missing):
if not k:
continue
l = sorted(missing[k])
n = len(l)
print w(
repr(tuple(int(y) for y in sorted(k))) + ': %i samples' % n)
for i in xrange(0, n, 6):
for j in xrange(i, min(i + 6, n)):
print w(' ' + l[j]),
print
warning()
def draw():
ps = plot_saver(plot_dir('o2t_templates_run2'), size=(600,600))
f = ROOT.TFile(gp.fn)
whiches = [
('multijet', [
(name2isample(f, 'mfv_neu_tau00100um_M0800'), ROOT.kRed, 'multijet M = 800 GeV, #tau = 100 #mum'),
(name2isample(f, 'mfv_neu_tau00300um_M0800'), ROOT.kGreen+2, 'multijet M = 800 GeV, #tau = 300 #mum'),
(name2isample(f, 'mfv_neu_tau01000um_M0800'), ROOT.kBlue, 'multijet M = 800 GeV, #tau = 1 mm'),
(name2isample(f, 'mfv_neu_tau10000um_M0800'), ROOT.kMagenta, 'multijet M = 800 GeV, #tau = 10 mm'),
(name2isample(f, 'mfv_neu_tau30000um_M0800'), ROOT.kOrange+2, 'multijet M = 800 GeV, #tau = 30 mm'),
]),
('dijet', [
(name2isample(f, 'mfv_ddbar_tau00100um_M0800'), ROOT.kRed, 'dijet M = 800 GeV, #tau = 100 #mum'),
(name2isample(f, 'mfv_ddbar_tau00300um_M0800'), ROOT.kGreen+2, 'dijet M = 800 GeV, #tau = 300 #mum'),
(name2isample(f, 'mfv_ddbar_tau01000um_M0800'), ROOT.kBlue, 'dijet M = 800 GeV, #tau = 1 mm'),
(name2isample(f, 'mfv_ddbar_tau10000um_M0800'), ROOT.kMagenta, 'dijet M = 800 GeV, #tau = 10 mm'),
(name2isample(f, 'mfv_ddbar_tau30000um_M0800'), ROOT.kOrange+2, 'dijet M = 800 GeV, #tau = 30 mm'),
]),
]
def fmt(h, name, color, save=[]):
binning = to_array(0., 0.04, 0.07, 0.15)
h = h.Rebin(len(binning)-1, name, binning)
h.Sumw2()
h.SetStats(0)
h.SetLineWidth(3)
h.SetLineColor(color)
h.SetTitle(';d_{VV} (cm);event rate (unit norm.)')
move_overflow_into_last_bin(h)
h.Scale(1./h.Integral(0,h.GetNbinsX()+2))
save.append(h)
return h
for which_name, which in whiches:
hbkg = fmt(f.Get('h_bkg_dvv'), 'bkg', ROOT.kBlack)
hbkg.Draw('hist e')
hbkg.GetYaxis().SetRangeUser(0,1.5)
leg = ROOT.TLegend(0.142, 0.657, 0.702, 0.857)
leg.SetBorderSize(0)
leg.AddEntry(hbkg, 'Background template d_{VV}^{C}', 'LE')
for isample, color, title in which:
h = fmt(f.Get('h_signal_%i_dvv' % isample), title, color)
h.Draw('hist e same')
leg.AddEntry(h, title, 'LE')
leg.Draw()
ps.save(which_name, log=False)
def __init__(self,
data_fn,
mc_fn,
data_path='pileup',
mc_path='PileupDist/h_npu',
tol=1e-9,
raise_tol=True):
from JMTucker.Tools.ROOTTools import ROOT
self.data_f = ROOT.TFile(data_fn)
self.mc_f = ROOT.TFile(mc_fn)
self.data_h = self.data_f.Get(data_path)
self.mc_h = self.mc_f.Get(mc_path)
def norm(h):
h = h.Clone(h.GetName() + '_norm')
h.Sumw2()
h.Scale(1 / h.Integral(1, h.GetNbinsX() + 1))
return h
self.data_h_orig = self.data_h
self.mc_h_orig = self.mc_h
self.data_h = norm(self.data_h)
self.mc_h = norm(self.mc_h)
self.ndata = self.data_h.GetNbinsX()
assert self.ndata == self.mc_h.GetNbinsX() # JMTBAD
self.weights = []
for i in xrange(1, self.ndata + 1):
d = self.data_h.GetBinContent(i)
m = self.mc_h.GetBinContent(i)
w = -1
if m == 0:
if d > tol:
msg = 'm == 0 and d = %g > tol=%g for i = %i' % (d, tol, i)
if raise_tol:
raise ValueError(msg)
else:
print msg
w = 0
else:
w = d / m
self.weights.append(w)
while self.weights[-1] == 0:
self.weights.pop()
def points(f=None):
if f is None:
f = ROOT.TFile(gp.fn)
kinds = sorted(set(s.kind for s in sample_iterator(f)))
masses = sorted(set(s.mass for s in sample_iterator(f)))
taus = sorted(set(s.tau for s in sample_iterator(f)))
return kinds, masses, taus
def signal_efficiency():
from signal_efficiency import SignalEfficiencyCombiner
for which, years in ('run2', [2016, 2017, 2018]), ('2017p8', [2017, 2018]):
combiner = SignalEfficiencyCombiner(years)
in_f = combiner.inputs[0].f
out_f = ROOT.TFile('signal_efficiency_%s.root' % which, 'recreate')
kinds, masses, taus, nmasses, ntaus = axes(in_f)
taus.remove(30.)
ntaus -= 1
for kind in kinds:
h = ROOT.TH2D('signal_efficiency_%s' % kind,
';mass (GeV);#tau (mm)', nmasses, masses, ntaus,
taus)
for ibin in xrange(1, nmasses + 1):
mass = h.GetXaxis().GetBinLowEdge(ibin)
for jbin in xrange(1, ntaus + 1):
tau = h.GetYaxis().GetBinLowEdge(jbin)
pt = kind, tau, mass
isample, r = None, None
try:
isample = name2isample(in_f, details2name(*pt))
r = combiner.combine(isample)
except ValueError as exc:
if tau < 100:
print colors.warning(
'problem getting %r : isample %r exc %s' %
(pt, isample, exc))
continue # leave holes in the plot to know which samples are missing
h.SetBinContent(ibin, jbin, r.total_sig_eff)
h.SetBinError(ibin, jbin, 0) # JMTBAD
out_f.cd()
h.Write()
out_f.Write()
out_f.Close()
def draw(self, fn):
from JMTucker.Tools.ROOTTools import ROOT, draw_in_order, differentiate_stat_box
self.data_h.SetLineColor(ROOT.kBlack)
self.mc_h.SetLineColor(ROOT.kRed)
self.data_h.SetLineWidth(2)
self.mc_h.SetLineWidth(2)
c = ROOT.TCanvas('c', '', 1000, 800)
draw_in_order([[self.data_h, self.mc_h], ''], sames=True)
c.Update()
differentiate_stat_box(self.mc_h)
if '.' in os.path.basename(fn):
c.SaveAs(fn)
else:
c.SaveAs(fn + '.png')
c.SaveAs(fn + '.root')
def __call__(self, era, eta):
if era == '2017D' or era == '2017E':
era = '2017DE'
name = era + eta
if self.fcns.has_key(name):
return self.fcns[name]
offset = self.eras.index(era) * 4 + self.etas.index(eta) * 2
pars = [float(x) for x in self.lines[offset].split('{')[1].split('}')[0].split(',')]
fcn = self.lines[offset+1].split(' = ')[1].replace(';','').replace('p_dxy[', '[p')
assert '[p%i]' % (len(pars)-1) in fcn
fcn = self.fcns[name] = ROOT.TF1('fcn_' + name, fcn, 1, 200)
fcn.SetParameters(*pars)
if self.debug:
print era, eta, offset, self.lines[offset], self.lines[offset+1], len(pars), pars
return fcn
def fromtree(fn, _jobre=re.compile(r'_(\d+)\.root$')):
bn = os.path.basename(fn)
f = ROOT.TFile(fn)
ll = [x for x in detree(f.Get('limit'), 'limit:iToy', xform=(float, int))]
if bn.startswith('expected'):
mo = _jobre.search(bn)
if mo:
njobsexp = 1
ntoysexp = ntoysperjob
else:
njobsexp = njobs
ntoysexp = njobs * ntoysperjob
if len(
ll
) == 2 * ntoysexp: # for some reason, there are two entries for every iToy when --saveToys used
ll2 = []
for mi, (l, i) in enumerate(ll):
assert (mi / 2 % ntoysperjob) + 1 == i
if mi % 2 == 1:
ll2.append(l)
elif mi % 2 * ntoysperjob == 0: # and the limit value saved in the first entry for the first toy in each batch is ~1e-300!
assert l < 1e-300
elif mi > 0:
assert l == ll2[-1]
assert len(ll2) == ntoysexp
return ll2
else:
if len(ll) != ntoysexp:
raise ValueError('fromtree %s failed: len(ll) = %i != %i' %
(fn, len(ll), ntoysexp))
jobordercheck = [b for _, b in ll]
jobordershouldbe = range(1, ntoysperjob + 1) * njobsexp
if jobordercheck != jobordershouldbe:
raise ValueError('fromtree %s failed: jobordercheck %r != $r' %
(fn, jobordercheck, jobordershouldbe))
return [a for a, _ in ll]
else:
assert bn == 'observed.root' and len(ll) == 1 and ll[0][1] == 0
return [ll[0][0]]
h.Write()
out_f.Write()
out_f.Close()
if __name__ == '__main__':
if 'make' in sys.argv:
make()
elif 'uncert' in sys.argv:
sig_uncert_2017p8(sys.argv[sys.argv.index('uncert')+1], debug=True)
elif 'draw' in sys.argv:
draw()
elif 'compare' in sys.argv:
compare(*sys.argv[2:5])
elif 'ngen' in sys.argv:
f = ROOT.TFile(gp.fn)
years = sys.argv[sys.argv.index('ngen'):]
for s in sample_iterator(f, years, slices_1d=True):
print '%6i' % s.isample, s.name.ljust(35), ' '.join(['%10s']*gp.nyears) % tuple(ngen(f, s.isample, year) for year in gp.years)
#nevents_plot()
elif 'points' in sys.argv:
print 'kinds = %r\nmasses = %r\ntaus = %r' % points()
elif 'iter' in sys.argv:
include_2016 = 'include_2016' in sys.argv
years = ('2016','2017','2018') if include_2016 else ('2017','2018')
samples = sample_iterator(ROOT.TFile('limitsinput.root'),
require_years=years,
test='test_batch' in sys.argv,
slices_1d='slices_1d' in sys.argv,
)
names = set(s.name for s in samples)
def phi(v):
return atan2(v[1], v[0])
def dphi(v0, v1):
phi0 = phi(v0)
phi1 = phi(v1)
res = phi0 - phi1
while res > pi:
res -= 2*pi
while res <= -pi:
res += 2*pi
return res
h_y_x = []
for i in xrange(0,10):
h_y_x.append(ROOT.TH2F('svdist2d_%i'%i, '%i um <= svdist2d < %i um;x (cm);y (cm)'%(100*i,100*(i+1)), 200, -0.1, 0.1, 200, -0.1, 0.1))
h_dz = ROOT.TH1F('h_dz', '', 100, -0.2, 0.2)
h_dphi = ROOT.TH1F('h_phi', '', 100, -pi, pi)
h_y_x_g = [[] for x in xrange(10)]
h_dzs = [ROOT.TH1F('h_dz_%i' % x, '', 100, -0.2, 0.2) for x in xrange(10)]
h_dphis = [ROOT.TH1F('h_dphi_%i' % x, '', 100, -pi, pi) for x in xrange(10)]
for sample, ntracks01, svdist2d, v0, v1 in events:
svdist2d *= 10000
bin = int(svdist2d / 100)
#print svdist2d, bin, event[3][0]-0.244, event[3][1]-0.3928, event[4][0]-0.244, event[4][1]-0.3928
dz = v0[-1] - v1[-1]
v0 = debs(v0)
def make_h_fcn(nm, fcn):
h = ROOT.TH1D(nm, '%s - %s' % (fn1, fn2), nnames, 0, nnames)
h_fcns.append((h, fcn))
return h
def compare(fn1, fn2, outbase):
f1, f2 = fs = ROOT.TFile.Open(fn1), ROOT.TFile.Open(fn2)
samples_1 = dict((name, isample) for isample, name in isamplename_iterator(f1))
samples_2 = dict((name, isample) for isample, name in isamplename_iterator(f2))
names_1 = set(samples_1)
names_2 = set(samples_2)
names = sorted(names_1 & names_2)
nnames = len(names)
if not names:
print 'no samples in common between %s and %s' % (fn1, fn2)
return
else:
print '%i samples in common' % nnames
if names_1 != names_2:
print '# samples only in %s: %i' % (fn1, len(names_1 - names_2))
#pprint(sorted(names_1 - names_2))
print '# samples only in %s: %i' % (fn2, len(names_2 - names_1))
#pprint(sorted(names_2 - names_1))
h_fcns = []
def make_h_fcn(nm, fcn):
h = ROOT.TH1D(nm, '%s - %s' % (fn1, fn2), nnames, 0, nnames)
h_fcns.append((h, fcn))
return h
h_eff1v = make_h_fcn('eff1v', eff1v)
h_eff2v = make_h_fcn('eff2v', eff2v)
h_dbvmean = make_h_fcn('dbvmean', lambda f,isample: (h_dbv(f,isample).GetMean(), h_dbv(f,isample).GetMeanError()))
h_dbvrms = make_h_fcn('dbvrms', lambda f,isample: (h_dbv(f,isample).GetRMS(), h_dbv(f,isample).GetRMSError() ))
h_dvvmean = make_h_fcn('dvvmean', lambda f,isample: (h_dvv(f,isample).GetMean(), h_dvv(f,isample).GetMeanError()))
h_dvvrms = make_h_fcn('dvvrms', lambda f,isample: (h_dvv(f,isample).GetRMS(), h_dvv(f,isample).GetRMSError() ))
all_stat_same, all_same = True, True
for iname, name in enumerate(names):
isamples = samples_1[name], samples_2[name]
for h, fcn in h_fcns:
(v1,e1), (v2, e2) = [fcn(f, isample) for f,isample in zip(fs, isamples)]
h.GetXaxis().SetBinLabel(iname+1, name)
d = v1 - v2
e = (e1**2 + e2**2)**0.5
h.SetBinContent(iname+1, d)
h.SetBinError (iname+1, e)
if abs(d) > 1e-5:
all_same = False
if abs(d) > e:
all_stat_same = False
c = ROOT.TCanvas('c', '', 1000, 800)
line = ROOT.TLine(0, 0, nnames, 0)
line.SetLineStyle(2)
line.SetLineWidth(2)
for i,(h,_) in enumerate(h_fcns):
h.SetStats(0)
h.SetLineWidth(2)
h.Draw('e')
line.Draw()
for ext in 'root', 'png':
c.SaveAs(outbase + '/%i_%s.%s' % (i, h.GetName(), ext))
print 'all same? %r statistically? %r' % (all_same, all_stat_same)
#!/usr/bin/env python
import sys
from pprint import pprint
from JMTucker.Tools.ROOTTools import ROOT, flatten_directory, check_consistency
ROOT.TH1.AddDirectory(0)
fn1, fn2 = fns = [x for x in sys.argv[1:] if '.root' in x]
print 'comparing', fn1, fn2
f1, f2 = fs = [ROOT.TFile(fn) for fn in fns]
d1 = set(flatten_directory(f1))
d2 = set(flatten_directory(f2))
# could remap
#assert sorted(d1) == sorted(d2)
d = sorted(d1 & d2)
in1not2 = sorted(d1 - d2)
in2not1 = sorted(d2 - d1)
if 'small' in sys.argv:
print 'comparing %i hists (and not %i+%i of them)' % (len(d), len(in1not2),
len(in2not1))
elif in1not2 or in2not1:
print 'warning: not comparing these:'
print 'in f1 not f2:'
pprint(in1not2)
print 'in f2 not f1:'
pprint(in2not1)
print
def book(out_name, hs=[]):
bins = binning()
h = ROOT.TH1F(out_name, '', len(bins) - 1, bins)
hs.append(h)
return h
mass = float(mass)
tau, tauunit = float(tau_s[:-2]), tau_s[-2:]
if tauunit == 'um':
tau *= 1e-4
elif tauunit == 'mm':
tau *= 1e-1
nevents_expected = 10000
if model == 'GluinoGluinoToNeutralinoNeutralinoTo2T2B2S':
nevents_missing_allowed = 0
folder_name = 'Lsps'
elif model == 'StopStopTo2Dbar2Dbar':
nevents_missing_allowed = 10
folder_name = 'Hs'
f = ROOT.TFile(fn)
hs = [f.Get('mfvGenHistos/%s#%i/M' % (folder_name, i)) for i in 0, 1]
ok = all(nevents_expected - h.GetEntries() <= nevents_missing_allowed
and abs(h.GetMean() - mass) < 0.01 for h in hs)
h = f.Get('mfvGenHistos/h_ctaubig')
xmax = h.GetXaxis().GetXmax()
mean_expected = (tau - exp(-xmax / tau) *
(xmax + tau)) / (1 - exp(-xmax / tau))
ok = ok and nevents_expected - h.GetEntries(
) / 2. <= nevents_missing_allowed and abs(
h.GetMean() - mean_expected) < 2 * h.GetMeanError()
print(colors.green if ok else colors.red)(
'%s (%s %s %s) %s (%s %s) %s (%s)' %
(fn, hs[0].GetEntries(), hs[1].GetEntries(), h.GetEntries(), mass,
return 6
elif mass == 1200:
return 7
elif mass == 1600:
return 8
elif mass == 3000:
return 9
matched = []
not_matched = []
x = []
y = []
ex = []
ey = []
gs = []
l1 = ROOT.TLegend(0.75,0.1,0.95,0.5)
l2 = ROOT.TLegend(0.75,0.5,0.95,0.9)
for sample in samples:
if ctau not in sample.name:
continue
fn = os.path.join(os.path.expanduser(file_path), '%s.root' % sample.name)
if not os.path.isfile(fn):
print colors.red('%s is missing' % fn) + '\n'
continue
print sample.name
f = ROOT.TFile(fn)
nrec = f.Get('mfvTheoristRecipe%s/h_gen%s_dvv' % (rec_den, match)).GetEntries()
ngen = f.Get('mfvGen%s/h_gen%s_dvv' % (gen_den, match)).GetEntries()
def make_bkg(f):
# the first 3 sets of values override what's in 2v_from_jets
observed = (1,0,0), (0,0,0), (0,0,0)
bkg_n1v = 1183, 1303, 908
bkg_n2v = 1, 0.241, 0.111 # this is not supposed to be the sum of observed, but could/should be set to the pre-fit expectation (predict2v.py)
# but bkg_c1v is checked against what's in 2v_from_jets
bkg_c1v = (0.509, 0.374, 0.117), (0.709, 0.257, 0.034), (0.650, 0.313, 0.037)
bkg_uncert = [(1.25, 1.25, 1.69), (1.16, 1.30, 1.51), (1.19, 1.30, 1.91)] # combine lnN convention
import statmodel
bkg_uncert_stat = [statmodel.ebins['data100pc_%s_5track' % year] for year in gp.years]
for y in xrange(gp.nyears):
bkg_uncert[y] = [(a**2 + b**2)**0.5 for a,b in zip(bkg_uncert[y], bkg_uncert_stat[y])] # JMTBAD use proper gmN?
bkg_stat_uncert = [(1.071, 1.128, 1.465), (1.173, 1.216, 1.454), (1.217, 1.238, 1.586)] # stat errors uncorrelated
bkg_syst_uncert = [(0.758, 1.218, 1.542), (0.743, 1.338, 1.389), (0.766, 1.315, 1.760)] # syst error anti-correlated
def bkg_fn(year, which='default'):
path = '/uscms/home/tucker/public/mfv/'
if year == '2016':
return path + '2v_from_jets_data_2015p6_5track_default_v15_v5.root'
else:
if which == 'c':
which = 'btag_corrected_nom'
else:
assert which == 'default'
return path + '2v_from_jets_data_%s_5track_%s_V27m.root' % (year, which)
def bkg_f(year, which='default', _c={}):
k = year, which
if not _c.has_key(k):
_c[k] = ROOT.TFile.Open(bkg_fn(*k))
return _c[k]
def cbkg_f(year):
return bkg_f(year, which='c')
##
print 'make_bkg:',
for y, year in enumerate(gp.years):
h_int_lumi = ROOT.TH1D('h_int_lumi_%s' % year, '', 1, 0, 1)
h_int_lumi.SetBinContent(1, gp.int_lumis[y])
h_observed = ROOT.TH1D('h_observed_%s' % year, '', gp.nbins, gp.bins)
for i,v in enumerate(observed[y]):
h_observed.SetBinContent(i+1, v)
h_bkg_dbv = to_TH1D( bkg_f(year).Get('h_1v_dbv'), 'h_bkg_dbv_%s' % year)
h_bkg_dvv = to_TH1D(cbkg_f(year).Get('h_c1v_dvv'), 'h_bkg_dvv_%s' % year)
h_bkg_dbv.Scale(bkg_n1v[y]/get_integral(h_bkg_dbv)[0])
h_bkg_dvv.Scale(bkg_n2v[y]/get_integral(h_bkg_dvv)[0])
h = h_bkg_dvv_rebin = h_bkg_dvv.Rebin(gp.nbins, 'h_bkg_dvv_rebin_%s' % year, gp.bins)
move_overflow_into_last_bin(h_bkg_dvv_rebin)
for i in xrange(gp.nbins):
if abs(bkg_c1v[y][i] - h.GetBinContent(i+1)/bkg_n2v[y]) > 0.001:
print y,i, bkg_c1v[y][i], h.GetBinContent(i+1)/bkg_n1v[y]
assert 0
h_bkg_uncert = ROOT.TH1D('h_bkg_uncert_%s' % year, '', gp.nbins, gp.bins)
h_bkg_stat_uncert = ROOT.TH1D('h_bkg_uncert_stat_%s' % year, '', gp.nbins, gp.bins)
h_bkg_syst_uncert = ROOT.TH1D('h_bkg_uncert_syst_%s' % year, '', gp.nbins, gp.bins)
for i,v in enumerate(bkg_uncert[y]):
h_bkg_uncert.SetBinContent(i+1, v)
for i,v in enumerate(bkg_stat_uncert[y]):
h_bkg_stat_uncert.SetBinContent(i+1, v)
for i,v in enumerate(bkg_syst_uncert[y]):
h_bkg_syst_uncert.SetBinContent(i+1, v)
f.cd()
for h in h_int_lumi, h_observed, h_bkg_dbv, h_bkg_dvv, h_bkg_dvv_rebin, h_bkg_uncert, h_bkg_stat_uncert, h_bkg_syst_uncert:
h.SetTitle('')
h.Write()
print 'done'
def make_signals_2015p6(f, name_list):
# Pull 2015+6 values from previous-format limitsinput file(s).
# (Not all 2017+8 signal points exist in the final one for the
# 2016 paper, but we make new ones now using the scanpack2017p8_2016 branch.)
# Rescale+add how SignalEfficiencyCombiner did previously. Go
# ahead and store the rate already scaled by the integrated
# luminosity. This makes the 2016 weighted sum situation easier
# downstream.
fnpairs = [ # (hip, nonhip)--the first pair are to be the final 2016 paper one, the rest define newly generated signal points
('/uscms/home/tucker/public/mfv/limitsinput_data_v15v5_scanpack_merge_hip_1_2_2p6_3_3p6_removeddbar.root',
'/uscms/home/tucker/public/mfv/limitsinput_data_v15v5_scanpack_merge_1_1p5_2_2p5_2p7_3_3p5_removeddbar.root',),
('/uscms/home/tucker/public/mfv/limitsinput_scanpack1D2016missing_hip.root',
'/uscms/home/tucker/public/mfv/limitsinput_scanpack1D2016missing.root',),
('/uscms/home/tucker/public/mfv/limitsinput_scanpack4p6_hip.root',
'/uscms/home/tucker/public/mfv/limitsinput_scanpack4p6.root',),
]
def trigmult2016(x):
return 1 # this was 0.99 in signal_efficiency, but now it's already in int_lumi at top
def sf20156(x, pars=(0.9784, -1128., 1444.)):
return trigmult2016(x) * (2. - pars[0] * ROOT.TMath.Erf((x-pars[1])/pars[2]))
sfs = [sf20156, trigmult2016, trigmult2016]
int_lumis = [2592.4, 19492., 16059.]
for ipair, (fn_2016_hip, fn_2016_nonhip) in enumerate(fnpairs):
name_list['title'] = name_list.get('title', '') + '+%s+%s' % (fn_2016_hip, fn_2016_nonhip)
f_2015 = f_2016 = f_2016_nonhip = ROOT.TFile.Open(fn_2016_nonhip)
f_2016_hip = ROOT.TFile.Open(fn_2016_hip)
h_name_list_2016 = f_2016_nonhip.Get('name_list') # already checked that the two files are in sync
def _hs(n):
return [f_2015.Get(n), f_2016_hip.Get(n), f_2016_nonhip.Get(n)]
nsigs = h_name_list_2016.GetNbinsX()
for ibin in xrange(1, nsigs+1):
if (ibin-1)%(nsigs/20) == 0:
sys.stdout.write('\rmake_signals_2015p6: pair %i/%i, %i/%i' % (ipair+1, len(fnpairs), ibin, nsigs)); sys.stdout.flush()
old_isample = -ibin
new_isample = old_isample if ipair == 0 else signalset.next_isample()
old_bn = 'h_signal_%i_' % old_isample
new_bn = 'h_signal_%i_' % new_isample
# convert sample name to new format--really only more digits in tau field
old_name = f_2016.Get(old_bn + 'norm').GetTitle()
mass = name2mass(old_name) # needed in scalefactor too
new_name = details2name(name2kind(old_name), name2tau(old_name), mass)
if name_list.has_key(new_name):
print colors.warning('\nwarning: %s found in ipair %i with isample %i was found in previous ipair %s with isample %i, skipping' % (new_name, ipair, old_isample, name_list[new_name].ipair, name_list[new_name].isample))
continue
name_list[new_name] = ss = signalset(new_name, '2016', new_isample)
ss.ipair = ipair
norms = _hs(old_bn + 'norm')
ngens = [1e-3/h.GetBinContent(2) for h in norms]
ngentot = sum(ngens[1:]) # hip/nonhip are the only independent samples, 2015 is just 2016 that we rescale
for n in 'dbv', 'dphi', 'dvv', 'dvv_rebin':
hs = _hs(old_bn + n)
for h,ng,sf,il in zip(hs, ngens, sfs, int_lumis):
if n == 'dbv' or n == 'dvv':
h.Rebin(10) # did too many bins last time
h.Scale(sf(mass) * 1e-3 * il / ng)
h = hs.pop(0)
for h2 in hs:
h.Add(h2)
h.SetName(new_bn + n + '_2016')
h.SetTitle(new_name + '_2016')
f.cd(); h.Write()
f.cd()
h = ROOT.TH1D(new_bn + 'uncert_2016', new_name + '_2016', gp.nbins, gp.bins)
uncerts = [0.24, sig_uncert_pdf(new_name + '_2016')] # ignore old hist, just use the previous total 24%
for ib in xrange(1,gp.nbins+1): h.SetBinContent(ib, 1+sum(x**2 for x in uncerts)**0.5)
h.Write()
h = ROOT.TH1D(new_bn + 'ngen_2016', new_name + '_2016', 1,0,1)
h.SetBinContent(1, ngentot)
h.Write()
print '\rmake_signals_2015p6: done with pair %i ' % (ipair+1)
import sys
from JMTucker.Tools.ROOTTools import ROOT
from binning import binning
def book(out_name, hs=[]):
bins = binning()
h = ROOT.TH1F(out_name, '', len(bins) - 1, bins)
hs.append(h)
return h
f = ROOT.TFile('export.root', 'recreate')
execfile(sys.argv[1])
f.Write()
f.Close()
def make_signals_2017p8(f, name_list):
# 2017,8 are from minitrees (the 100kevt official samples) and scanpack.
scanpack_list = '/uscms/home/tucker/public/mfv/scanpacks/2017p8/scanpack1D_4_4p7_4p8.merged.list.gz'
trees = '/uscms_data/d2/tucker/crab_dirs/MiniTreeV27m/mfv*.root'
title = []
sigs = {}
if scanpack_list:
title.append(scanpack_list)
sigs.update(eval(GzipFile(scanpack_list).read()))
if trees:
title.append(trees)
sigs.update({os.path.basename(fn).replace('.root', '') : [fn] for fn in sorted(glob(trees))}) # overrides scanpack entries
name_list['title'] = name_list.get('title', '') + '+' + '+'.join(title)
nsigs = len(sigs)
for isig, (name_year, fns) in enumerate(sigs.iteritems()):
if isig%(nsigs/20) == 0:
sys.stdout.write('\rmake_signals_2017p8: %i/%i' % (isig, nsigs)); sys.stdout.flush()
name, year = name_year.rsplit('_',1)
if name_list.has_key(name):
s = name_list[name]
s.years.add(year)
else:
s = name_list[name] = signalset(name, year)
n = lambda x: 'h_signal_%i_%s_%s' % (s.isample, x, year)
ngen = 0.
t = ROOT.TChain('mfvMiniTree/t')
for fn in fns:
sig_f = ROOT.TFile.Open(fn)
ngen += sig_f.Get('mfvWeight/h_sums').GetBinContent(1)
sig_f.Close()
t.Add(fn)
if year == '2017' and gp.l1eeprefiring_2017:
t.SetAlias('jet_l1ee', 'abs(jet_eta) > 2.25 && jet_pt > 100')
t.SetAlias('njets_l1ee', 'Sum$(!jet_l1ee)')
t.SetAlias('jetht_l1ee', 'Sum$(jet_pt * (jet_pt > 40 && !jet_l1ee))')
t.SetAlias('limitsinput_pass', '(njets_l1ee >= 4 && jetht_l1ee >= 1200)')
elif year == '2018' and gp.hem1516_2018:
t.SetAlias('jet_hem1516', 'jet_eta < -1.3 && jet_phi < -0.87 && jet_phi > -1.57')
t.SetAlias('njets_hem1516', 'Sum$(!jet_hem1516)')
t.SetAlias('jetht_hem1516', 'Sum$(jet_pt * (jet_pt > 40 && !jet_hem1516))')
t.SetAlias('limitsinput_pass', '(njets_hem1516 >= 4 && jetht_hem1516 >= 1200) || (rndm < 0.36)') # silly way to represent effect only on 64% of the data
else:
t.SetAlias('limitsinput_pass', '1==1')
iyear = gp.years.index(year)
scale = 1e-3 * gp.int_lumis[iyear] / ngen
data_mc_scale = sig_datamcSF_2017p8(name_year)
ROOT.TH1.AddDirectory(1) # the Draw>> output goes off into the ether without this stupid crap
h_dbv = ROOT.TH1D(n('dbv'), '', 1250, 0, 2.5)
h_dvv = ROOT.TH1D(n('dvv'), '', 400, 0, 4)
h_dphi = ROOT.TH1D(n('dphi'), '', 10, -3.15, 3.15)
t.Draw('dist0>>%s' % n('dbv'), 'weight*(limitsinput_pass && nvtx==1)')
t.Draw('svdist>>%s' % n('dvv'), 'weight*(limitsinput_pass && nvtx>=2)')
t.Draw('svdphi>>%s' % n('dphi'), 'weight*(limitsinput_pass && nvtx>=2)')
ROOT.TH1.AddDirectory(0)
for h in h_dbv, h_dvv, h_dphi:
h.SetDirectory(0)
h.Scale(scale)
if 'dbv' in h.GetTitle():
h.Scale(data_mc_scale)
else:
h.Scale(data_mc_scale**2)
h_dvv_rebin = h_dvv.Rebin(gp.nbins, n('dvv_rebin'), gp.bins)
move_overflow_into_last_bin(h_dvv_rebin)
h_ngen = ROOT.TH1D(n('ngen'), '', 1,0,1)
h_ngen.SetBinContent(1, ngen)
h_uncert = ROOT.TH1D(n('uncert'), '', gp.nbins, gp.bins)
for i,v in enumerate(sig_uncert_2017p8(name_year)):
h_uncert.SetBinContent(i+1, v)
f.cd()
for h in h_dbv, h_dphi, h_dvv, h_dvv_rebin, h_uncert, h_ngen:
h.SetTitle(name_year)
h.Write()
print '\rmake_signals_2017p8: done '
def cutplayMangle(filename):
debug = False
file = ROOT.TFile.Open(filename)
if not file.IsOpen():
raise RuntimeError('could not open %s' % filename)
norm = norm_from_file(filename)
x = file.Get('SimpleTriggerEfficiency/triggers_pass_num')
cutscans = {}
nm1 = None
# Extract from the big list of bins the numbers separately for each
# cut.
for i in xrange(1, x.GetNbinsX() + 1):
label = x.GetXaxis().GetBinLabel(i)
if label == 'nm1':
nm1 = x.GetBinContent(i), x.GetBinError(i)
continue
try:
cut, cut_val = label.split('X')
except ValueError:
print 'warning: could not parse label %s' % label
continue
if not cutscans.has_key(cut):
cutscans[cut] = []
cutscans[cut].append((cut_val, x.GetBinContent(i)))
h_intot = file.Get('SimpleTriggerEfficiency/triggers_pass_den')
ntot = h_intot.GetBinContent(1), h_intot.GetBinError(1)
file.Close()
output_fn = filename.replace('.root', '_mangled.root')
output_file = ROOT.TFile(output_fn, 'RECREATE')
sample_name = os.path.basename(filename).replace('.root', '')
try:
sample = getattr(Samples, sample_name)
except AttributeError:
sample = None
h_ntot = ROOT.TH1F('ntot', '', 1, 0, 1)
scale = ntot[0] / norm if sample is not None else 1.
h_ntot.SetBinContent(1, ntot[0] / scale)
h_ntot.SetBinError(1, ntot[1] / scale)
if nm1 is not None:
h_nm1 = ROOT.TH1F('nm1', '', 1, 0, 1)
h_nm1.SetBinContent(1, nm1[0])
h_nm1.SetBinError(1, nm1[1])
else:
raise ValueError('did not find nm1 value in input')
output_hists = []
for cut_name, scan in cutscans.iteritems():
if debug:
print 'cut: %s' % cut_name
hmin = depize(scan[0][0])
hminplus1 = depize(scan[1][0])
binwidth = hminplus1 - hmin
hmaxminus1 = depize(scan[-1][0])
hmax = hmaxminus1 + binwidth
if debug:
print 'hmin: ', hmin, ' hminplus1: ', hminplus1, ' binwidth: ', binwidth, ' hmaxminus1: ', hmaxminus1, ' hmax: ', hmax
hist = ROOT.TH1F(cut_name, '', len(scan), hmin, hmax)
output_hists.append(hist)
for cut_bin_name, nevents in scan:
bin = hist.FindBin(depize(cut_bin_name) + 0.00001)
bin_edge = hist.GetBinLowEdge(bin)
if debug:
print 'cut bin name: %s ibin: %i bin_edge: %s number of events: %f' % (
cut_bin_name, bin, bin_edge, nevents)
hist.SetBinContent(bin, nevents)
output_file.Write()
output_file.Close()
out_f.Write()
out_f.Close()
if __name__ == '__main__':
if 'make' in sys.argv:
make()
elif 'uncert' in sys.argv:
sig_uncert_2017p8(sys.argv[sys.argv.index('uncert') + 1], debug=True)
elif 'draw' in sys.argv:
draw()
elif 'compare' in sys.argv:
compare(*sys.argv[2:5])
elif 'ngen' in sys.argv:
f = ROOT.TFile(gp.fn)
years = sys.argv[sys.argv.index('ngen'):]
for s in sample_iterator(f, years, slices_1d=True):
print '%6i' % s.isample, s.name.ljust(35), ' '.join(
['%10s'] * gp.nyears) % tuple(
ngen(f, s.isample, year) for year in gp.years)
#nevents_plot()
elif 'points' in sys.argv:
print 'kinds = %r\nmasses = %r\ntaus = %r' % points()
elif 'iter' in sys.argv:
include_2016 = 'include_2016' in sys.argv
years = ('2016', '2017', '2018') if include_2016 else ('2017', '2018')
samples = sample_iterator(
ROOT.TFile('limitsinput.root'),
require_years=years,
test='test_batch' in sys.argv,
