# var : [histograms, var-range, matplotlib title, ROOT title]
'time': [[], (0, 15), 'time [ps]', 'time [ps]'],
'terr': [[], (0, 0.12), 'time error ($\delta t$) [ps]', 'time error (#delta t) [ps]'],
'lab0_MM': [[], (4700, 5700), r'$B_{s}$ mass [MeV]', 'B_{s} mass [MeV]']
}
## make efficiency histograms
from ROOT import TH1D, TCanvas
tmp = TCanvas('c', '', 400, 400) # temporary canvas
from utils import scan_range, make_varefffn, colours
from rplot.utils import th1offset
for var in variables:
htemp = TH1D('htemp', '', 100, *variables[var][1])
# iterates over mva_cuts, returns list of efficiency histograms
variables[var][0] = scan_range(make_varefffn(htemp, refcut),
mva_cuts, classifier, tree, var)
del htemp
# aesthetics
for j in xrange(len(mva_cuts)):
heff = variables[var][0][j]
if j > 0: th1offset(heff, j)
heff.SetYTitle('Efficiency (w/ offset)')
heff.SetXTitle(variables[var][3])
heff.SetLineColor(colours(j))
heff.SetMarkerColor(colours(j))
heff.SetMarkerSize(0.2)
heff.SetMaximum(3.5)
heff.SetMinimum(0)
heff.SetStats(False)
del tmp
nevts_passed += [
lambda tree, cut: tree.GetEntries('{}&&{}'.format(region1, cut.ge)),
lambda tree, cut: tree.GetEntries('{}&&{}'.format(region2, cut.ge)),
lambda tree, cut: tree.GetEntries('{}&&{}'.format(region3, cut.ge))
]
from numpy import linspace, array, isnan
if classifier == 'BDTB':
clrange = (-0.3, 0.3)
else:
clrange = (-1.0, 1.0)
cuts = linspace(clrange[0], clrange[1], 101)
# significance, signal selection and background rejection efficiency
from utils import scan_range
res = array(scan_range(nevts_passed, cuts, classifier, tree))
# NOTE: filter runtime warning due to NaNs. These entries are
# filtered away before filling the histograms. They occur due to 0
# efficiency for very large classifier cuts.
import warnings
warnings.filterwarnings(action='ignore', category=RuntimeWarning,
message='invalid value encountered in divide.*')
from math import sqrt
if istmva:
if eff:
# sel_s/tot_s, 1-sel_b/tot_s
effs = array([0, 1]) + res / array([nsig, -nbkg])
eff_s, eff_b = effs[:, 0], effs[:, 1]
nevts_passed += [
lambda tree, cut: tree.GetEntries('{}&&{}'.format(region1, cut.ge)),
lambda tree, cut: tree.GetEntries('{}&&{}'.format(region2, cut.ge)),
lambda tree, cut: tree.GetEntries('{}&&{}'.format(region3, cut.ge))
]
from numpy import linspace, array, isnan
if classifier == 'BDTB':
clrange = (-0.3, 0.3)
else:
clrange = (-1.0, 1.0)
cuts = linspace(clrange[0], clrange[1], 101)
# significance, signal selection and background rejection efficiency
from utils import scan_range
res = array(scan_range(nevts_passed, cuts, classifier, tree))
# NOTE: filter runtime warning due to NaNs. These entries are
# filtered away before filling the histograms. They occur due to 0
# efficiency for very large classifier cuts.
import warnings
warnings.filterwarnings(action='ignore',
category=RuntimeWarning,
message='invalid value encountered in divide.*')
from math import sqrt
if istmva:
if eff:
# sel_s/tot_s, 1-sel_b/tot_s
effs = array([0, 1]) + res / array([nsig, -nbkg])
eff_s, eff_b = effs[:, 0], effs[:, 1]
