def draw(s, variables):
xarr1 = read_tree(t_in1, variables + [weight1[s]], cut1[s])
xarr2 = read_tree(t_in2, variables + [weight2[s]], cut2[s])
ret = {}
for var in variables:
h1 = hbase[var].Clone(s + '_1')
h2 = hbase[var].Clone(s + '_2')
draw_hist(h1, xarr1, [var], weight1[s])
draw_hist(h2, xarr2, [var], weight2[s])
# h1.SetMaximum(h1.GetMaximum()*2)
# h2.SetMaximum(h2.GetMaximum()*2)
ret[var] = {1: h1, 2: h2}
return ret
def tree_to_arrays(infilepath, treename='inputs'):
f = root.TFile(infilepath)
t = f.Get(treename)
data = {}
singletons = r.read_tree(t, branches=singleton_branches)
for k in singleton_branches:
data[k] = singletons[k]
arr = r.read_tree(t, branches=['kinematics'])
data['pf'] = np.array([x[0].tolist() for x in arr])
arr = r.read_tree(t, branches=['svs'])
data['sv'] = np.array([x[0].tolist() for x in arr])
return data
def draw_all():
f_in = root.TFile.Open('merged.root')
t_in = f_in.Get('events')
f_out = root.TFile.Open('hists.root', 'RECREATE')
weight_strs = []
for idx in xrange(len(weights)):
if idx == 0:
weight_str = 'normalizedWeight'
else:
weight_str = 'normalizedWeight*fabs(weights[%i])' % (idx - 1)
weight_strs.append(weight_str)
n_weights = len(weight_strs)
n_per = 50
nominal_arr = None
for iw in xrange(0, n_weights, n_per):
PInfo(sname, 'Extracting %i -> %i' % (iw, iw + n_per))
xarr = read_tree(t_in, ['genBosonPt'] + weight_strs[iw:iw + n_per])
for idx in xrange(iw, iw + n_per):
if idx == n_weights:
break
h = hbase.Clone()
weight_str = weight_strs[idx]
weight_name = weights[idx]
draw_hist(h, xarr, ['genBosonPt'], weight_str)
f_out.WriteTObject(h, 'h_' + weight_name)
f_out.Close()
f_in.Close()
def convert(tree,
branches=['px', 'py', 'pz', 'e', 'ptype'],
astype=np.float16):
struct_arr = read_tree(tree, branches)
arr = np.stack([struct_arr[f].astype(astype) for f in branches],
axis=1).transpose((0, 2, 1))
return arr
def tree_to_arrays(infilepath, treename='inputs'):
f = root.TFile(infilepath)
t = f.Get(treename)
data = {}
singleton_branches = [
x.GetName() for x in t.GetListOfBranches()
if (x.GetName() != 'kinematics')
]
singleton_branches = sorted(list(set(singleton_branches)))
singletons = r.read_tree(t, branches=singleton_branches)
for k in singleton_branches:
data[k] = singletons[k]
data['singleton_branches'] = np.array(singleton_branches)
arr = r.read_tree(t, branches=['kinematics'])
data['particles'] = np.array([x[0].tolist() for x in arr])
return data
def run(self, f_out):
PInfo('fitting_forest.Process.run', 'Running ' + self.name)
branches = sorted(self.all_branches.values())
try:
xarr = root_interface.read_tree(tree=self.tree,
branches=branches,
cut=self.cut)
fields = self.variables.keys() + ['nominal']
self.__write_out(f_out, xarr, fields, '')
for shift, weight in self.weights.iteritems():
fields = self.variables.keys() + [shift]
self.__write_out(f_out, xarr, fields, '_' + shift)
except ValueError as e:
PError('fitting_forest.Process.run', str(e))
return
def build_hists(cut,weight,observable):
h = {}
weight = weight%1
cut = tAND(cut,add_cut)
weights = {
'13TeV' : weight,
'8TeV' : weight.replace('sf_tt','sf_tt8TeV'),
'none' : weight.replace('sf_tt','1')
}
xarr = read_tree(tree,[observable]+weights.values(),cut)
for label,weight in weights.iteritems():
h[label] = hbase.Clone(label+'_'+observable)
draw_hist(h[label],xarr,[observable],weight)
divide_bin_width(h[label])
return h
def draw(s):
weight_ = weight
cut_ = cut
to_plot_ = [x[0] for x in to_plot]
for k,v in repl[s].iteritems():
cut_ = cut_.replace(k,v)
for i in xrange(len(to_plot_)):
to_plot_[i] = to_plot_[i].replace(k,v)
cut_ = cut_.replace('dphipfmetUp','dphipfmet')
cut_ = cut_.replace('dphipfmetDown','dphipfmet')
xarr = read_tree(tree=t_in,branches=to_plot_+[weight_],cut=cut_)
ret = {}
for name,shifted_name in zip(to_plot,to_plot_):
h = hbase[name[0]].Clone(shifted_name)
draw_hist(h,xarr,[shifted_name],weight_)
ret[name[0]] = h
return ret
def draw_all():
f_in = root.TFile.Open('merged.root')
t_in = f_in.Get('events')
f_out = root.TFile.Open('hists.root', 'RECREATE')
weight_strs = []
for idx in xrange(len(weights)):
if idx == 0:
weight_str = 'normalizedWeight'
else:
weight_str = 'normalizedWeight*weights[%i]' % (idx - 1)
weight_strs.append(weight_str)
xarr = read_tree(t_in, ['genBosonPt'] + weight_strs)
for idx in xrange(len(weights)):
h = hbase.Clone()
weight_str = weight_strs[idx]
weight_name = weights[idx]
draw_hist(h, xarr, ['genBosonPt'], weight_str)
f_out.WriteTObject(h, 'h_' + weight_name)
f_out.Close()
f_in.Close()
e_num = hnum.GetBinError(ib)
v_num = hnum.GetBinContent(ib)
hratio.SetBinContent(ib,v_num/v_den)
hratio.SetBinError(ib,e_num/v_den)
return hratio
def build_hists(cut,weight,observable):
h = {}
weight = weight%1
cut = tAND(cut,add_cut)
weights = {
'13TeV' : weight,
'8TeV' : weight.replace('sf_tt','sf_tt8TeV'),
'none' : weight.replace('sf_tt','1')
}
xarr = read_tree(tree,[observable]+weights.values(),cut)
for label,weight in weights.iteritems():
h[label] = hbase.Clone(label+'_'+observable)
draw_hist(h[label],xarr,[observable],weight)
divide_bin_width(h[label])
return h
h_sig = build_hists(sel.cuts['signal'],sel.weights['signal'],'pfmet')
h_con = build_hists(sel.cuts['singlemuontop'],sel.weights['singlemuontop'],'pfUWmag')
for k in h_sig:
h_sig[k].Divide(h_con[k])
def interpolate():
f_interp_path = basedir + '/interpolate/hists_couplings/%s.root' % (
args.mass)
f_interp = root.TFile.Open(f_interp_path)
PInfo(sname, 'We want to interpolate point %s' % args.mass)
if not args.mass_reco:
m_V, m_DM = [int(x) for x in args.mass.split('_')]
offshell = (2 * m_DM >= m_V)
frescos = map(
lambda x: x.split('/')[-1].replace('.root', '').replace(
'fittingForest_signal_vector_', ''),
glob(basedir + '/signals_3/fittingForest_signal_vector*root'))
frescos = map(lambda x: tuple(map(lambda y: int(y), x.split('_'))),
frescos)
m_V_recos = [x[0] for x in frescos]
m_V_reco = min(m_V_recos, key=lambda x: abs(m_V - x))
m_DM_recos = [x[1] for x in frescos if x[0] == m_V_reco]
m_DM_recos = [x for x in m_DM_recos if ((2 * x >= m_V) == offshell)]
if len(m_DM_recos) == 0:
m_DM_recos = [x[1] for x in frescos if x[0] == m_V_reco]
m_DM_reco = min(m_DM_recos, key=lambda x: abs(m_DM - x))
args.mass_reco = '%i_%i' % (m_V_reco, m_DM_reco)
PInfo(sname,
'We have chosen as the reconstructed point %s' % args.mass_reco)
f_reco_path = basedir + '/signals_3/fittingForest_signal_vector_%s.root' % (
args.mass_reco)
f_reco = root.TFile.Open(f_reco_path)
t_reco = f_reco.Get('%s_signal' % (args.mass_reco))
f_reco_gen_path = basedir + '/interpolate/hists/%s.root' % (args.mass_reco)
f_reco_gen = root.TFile.Open(f_reco_gen_path)
system('mkdir -p %s/interpolate/interpolated_couplings2test/' % basedir)
t_out = t_reco.Clone()
metname = 'min(met,999.9999)'
branches = [metname]
interp2reco = {}
for g in couplings_interp:
# find the reco g
# g_reco = min(couplings_reco, key = lambda x : euclid(parse_coupling_name(g), parse_coupling_name(x)))
g_reco = 'nominal' # override
PInfo(sname, 'Interpolating %s with %s' % (g, g_reco))
interp2reco[g] = g_reco
h_interp = f_interp.Get('h_%s' % g)
h_reco_gen = f_reco_gen.Get('h_%s' % g_reco)
h_ratio = h_interp.Clone()
h_ratio.Divide(h_reco_gen)
ba.newBranchName = 'interp_%s' % g
ba.AddBranchFromHistogram(t_reco, h_ratio)
if g_reco == 'nominal':
weightname = 'weight*interp_%s' % g
else:
weightname = 'weight*%s*interp_%s' % (g_reco.replace(
'_nlo', '').replace('rw_', ''), g)
branches.append(weightname)
for s in systs:
branches.append(weightname.replace('weight', s))
xarr = {}
xarr['tight'] = read_tree(t_reco, branches, 'top_ecf_bdt>0.45 && met>250')
xarr['loose'] = read_tree(
t_reco, branches, 'top_ecf_bdt>0.1 && top_ecf_bdt<0.45 && met>250')
f_out = root.TFile(
'%s/interpolate/interpolated_couplings2test/interp_%s.root' %
(basedir, args.mass), 'RECREATE')
PInfo(
sname,
'Creating %s/interpolate/interpolated_couplings2test/interp_%s.root' %
(basedir, args.mass))
for g in couplings_interp:
f_out.cd()
folder = f_out.mkdir(g)
h_templates = {}
g_reco = interp2reco[g]
if g_reco == 'nominal':
weightname = 'weight*interp_%s' % g
else:
weightname = 'weight*%s*interp_%s' % (g_reco.replace(
'_nlo', '').replace('rw_', ''), g)
for t in ['tight', 'loose']:
h_templates[t] = h_base.Clone()
draw_hist(h_templates[t], xarr[t], [metname], weightname)
for s in systs:
hname = '%s_%s' % (t, s)
h_templates[hname] = h_base.Clone()
draw_hist(h_templates[hname], xarr[t], [metname],
weightname.replace('weight', s))
for k, h in h_templates.iteritems():
folder.WriteTObject(h, 'h_%s' % (k))
f_out.Close()
def draw(tree,weight,cut):
h = hbase.Clone()
xarr = read_tree(tree,['met',weight],cut)
draw_hist(h,xarr,['met'],weight)
return h
def get_hist(cut):
h = hbase.Clone()
xarr = read_tree(events,['pfmetnomu'],cut)
draw_hist(h,xarr,['pfmetnomu'],None)
return h
