def read(scan, param, files, chop, remove_near_min, rezero,
remove_delta=None, improve=False, remove_dups=True):
# print files
goodfiles = [f for f in files if plot.TFileIsGood(f)]
limit = plot.MakeTChain(goodfiles, 'limit')
graph = plot.TGraphFromTree(
limit, param, '2*%s' % DELTANLL, 'quantileExpected > -1.5')
# print 'INPUT'
# graph.Print()
graph.SetName(scan)
graph.Sort()
if remove_dups:
plot.RemoveGraphXDuplicates(graph)
if remove_delta is not None:
plot.RemoveSmallDelta(graph, remove_delta)
plot.RemoveGraphYAbove(graph, chop)
plot.ReZeroTGraph(graph, rezero)
if remove_near_min is not None:
plot.RemoveNearMin(graph, remove_near_min)
if improve:
global NAMECOUNTER
spline = ROOT.TSpline3("spline3", graph)
func = ROOT.TF1('splinefn' + str(NAMECOUNTER), partial(Eval, spline),
graph.GetX()[0], graph.GetX()[graph.GetN() - 1], 1)
func.SetNpx(NPX)
NAMECOUNTER += 1
plot.ImproveMinimum(graph, func, True)
# graph.Print()
if FILTER is not None:
plot.FilterGraph(graph, FILTER)
if REMOVE_X_RANGES is not None:
for remove_x in REMOVE_X_RANGES:
plot.RemoveInXRange(graph, remove_x[0], remove_x[1])
return graph
def read(scan,
param,
files,
chop,
remove_near_min,
rezero,
remove_delta=None,
improve=False):
# print files
goodfiles = [f for f in files if plot.TFileIsGood(f)]
limit = plot.MakeTChain(goodfiles, 'limit')
# require quantileExpected > -0.5 to avoid the final point which is always committed twice
# (even if the fit fails)
graph = plot.TGraphFromTree(limit, param, '2*deltaNLL',
'quantileExpected > -0.5')
graph.SetName(scan)
graph.Sort()
plot.RemoveGraphXDuplicates(graph)
if remove_delta is not None: plot.RemoveSmallDelta(graph, remove_delta)
plot.RemoveGraphYAbove(graph, chop)
plot.ReZeroTGraph(graph, rezero)
if remove_near_min is not None: plot.RemoveNearMin(graph, remove_near_min)
if improve:
global NAMECOUNTER
spline = ROOT.TSpline3("spline3", graph)
func = ROOT.TF1('splinefn' + str(NAMECOUNTER), partial(Eval, spline),
graph.GetX()[0],
graph.GetX()[graph.GetN() - 1], 1)
NAMECOUNTER += 1
plot.ImproveMinimum(graph, func, True)
# graph.Print()
return graph
def main(args):
# Get tree with scan values from all input files
limit = plot.MakeTChain(args.files, 'limit')
# Write Tree entries in TGraph to get structure for likelihood database
graph = plot.TGraph2DFromTree(
limit, args.x_var, args.y_var, 'deltaNLL',
'quantileExpected > -0.5 && deltaNLL < 1000')
# 'quantileExpected > -0.5 && deltaNLL > 0 && deltaNLL < 1000')
# rezero the TGraph to have sensible values in the output file
min_delta_nll = rezero_tgraph2d(graph)
df = convert_graph_to_dataframe(graph)
df.to_csv(args.output,
sep=" ",
float_format="%.6f",
header=False,
index=False,
na_rep="NaN",
columns=[args.x_var, args.y_var, "deltaNLL"])
best = plot.TGraphFromTree(limit, args.x_var, args.y_var, 'deltaNLL == 0')
plot.RemoveGraphXDuplicates(best)
# Write back a root TTree with the modified contents.
outfile = ROOT.TFile(args.output.replace(".txt", ".root"), "recreate")
tree = convert_dataframe_to_tree(df, best, offset=min_delta_nll)
# Add the best fit values from the limit trees
outfile.Write()
return
def ProcessEnvelopeNew(main, others, relax_safety=0, chop=-1.):
print '[ProcessEnvelope] Will create envelope from %i other scans' % len(
others)
min_x = min([oth['graph'].GetX()[0] for oth in others])
max_x = max(
[oth['graph'].GetX()[oth['graph'].GetN() - 1] for oth in others])
# print '(min_x,max_x) = (%f, %f)' % (min_x, max_x)
npoints = 200
step = (max_x - min_x) / float(npoints - 1)
x = min_x
xvals = []
yvals = []
for i in xrange(npoints):
yset = []
for oth in others:
gr = oth['graph']
if x >= gr.GetX()[0] and x <= (gr.GetX()[gr.GetN() - 1] + 1E-6):
yset.append(oth['func'].Eval(x))
# print 'At x=%f, possible y vals are %s' % (x, yset)
if len(yset) > 0:
xvals.append(x)
yvals.append(min(yset))
x = x + step
gr = ROOT.TGraph()
gr.Set(len(xvals)) # will not contain the best fit
for i in xrange(gr.GetN()):
gr.SetPoint(i, xvals[i], yvals[i])
# print 'Envelope'
# gr.Print()
spline = ROOT.TSpline3("spline3", gr)
global NAMECOUNTER
func = ROOT.TF1('splinefn' + str(NAMECOUNTER), partial(Eval, spline),
gr.GetX()[0],
gr.GetX()[gr.GetN() - 1], 1)
func.SetNpx(NPX)
min_x, min_y = plot.ImproveMinimum(gr, func)
gr.Set(len(xvals) + 1)
gr.SetPoint(len(xvals), min_x, min_y)
gr.Sort()
for i in xrange(gr.GetN()):
gr.GetY()[i] -= min_y
if chop > 0:
plot.RemoveGraphYAbove(gr, chop)
for oth in others:
for i in xrange(oth['graph'].GetN()):
oth['graph'].GetY()[i] -= min_y
if chop > 0:
plot.RemoveGraphYAbove(oth['graph'], chop)
# print 'OTHER'
# oth['graph'].Print()
plot.RemoveGraphXDuplicates(gr)
return gr
def read(scan, param, files, ycut):
goodfiles = [f for f in files if plot.TFileIsGood(f)]
limit = plot.MakeTChain(goodfiles, 'limit')
graph = plot.TGraphFromTree(limit, param, '2*deltaNLL', 'quantileExpected > -1.5')
graph.SetName(scan)
graph.Sort()
plot.RemoveGraphXDuplicates(graph)
plot.RemoveGraphYAbove(graph, ycut)
# graph.Print()
return graph
def ProcessEnvelope(main, others, relax_safety=0):
'[ProcessEnvelope] Will create envelope from %i other scans' % len(others)
vals = {}
for oth in others:
gr = oth['graph']
# gr.Print()
for i in xrange(gr.GetN()):
x = gr.GetX()[i]
y = gr.GetY()[i]
if x not in vals:
vals[x] = []
vals[x].append(y)
lengths = []
for key in sorted(vals):
# print '%f %i' % (key,len(vals[key]))
lengths.append(len(vals[key]))
mode = max(set(lengths), key=lengths.count)
to_del = []
for key in sorted(vals):
if len(vals[key]) < (mode - relax_safety):
to_del.append(key)
for x in to_del:
del vals[x]
gr = ROOT.TGraph()
gr.Set(len(vals)) # will not contain the best fit
for i, key in enumerate(sorted(vals)):
gr.SetPoint(i, key, min(vals[key]))
# print 'Envelope'
# gr.Print()
spline = ROOT.TSpline3("spline3", gr)
global NAMECOUNTER
func = ROOT.TF1('splinefn' + str(NAMECOUNTER), partial(Eval, spline),
gr.GetX()[0],
gr.GetX()[gr.GetN() - 1], 1)
func.SetNpx(NPX)
min_x, min_y = plot.ImproveMinimum(gr, func)
gr.Set(len(vals) + 1)
gr.SetPoint(len(vals), min_x, min_y)
gr.Sort()
for i in xrange(gr.GetN()):
gr.GetY()[i] -= min_y
for oth in others:
for i in xrange(oth['graph'].GetN()):
oth['graph'].GetY()[i] -= min_y
# print 'OTHER'
# oth['graph'].Print()
plot.RemoveGraphXDuplicates(gr)
return gr
def read(scan, param_x, param_y, file):
# print files
goodfiles = [f for f in [file] if plot.TFileIsGood(f)]
limit = plot.MakeTChain(goodfiles, 'limit')
graph = plot.TGraph2DFromTree(limit, param_x, param_y, '2*deltaNLL', 'quantileExpected > -0.5 && deltaNLL > 0')
best = plot.TGraphFromTree(limit, param_x, param_y, 'quantileExpected > -0.5 && deltaNLL == 0')
plot.RemoveGraphXDuplicates(best)
assert(best.GetN() == 1)
graph.SetName(scan)
best.SetName(scan+'_best')
# graph.Print()
return (graph, best)
def read(scan, param, other_param, files, remove_dups=True):
# print files
goodfiles = [f for f in files if plot.TFileIsGood(f)]
limit = plot.MakeTChain(goodfiles, 'limit')
graph = plot.TGraphFromTree(limit, param, other_param,
'quantileExpected > -0.5')
# print 'INPUT'
# graph.Print()
graph.SetName(scan)
graph.Sort()
if remove_dups:
plot.RemoveGraphXDuplicates(graph)
# graph.Print()
return graph
###plot.RemoveGraphXDuplicates(best2)
####hists2 = plot.TH2FromTGraph2D(graph2, method='BinCenterAligned')
###hists.append( plot.TH2FromTGraph2D(graph2, method='BinCenterAligned') )
####plot.fastFillTH2(hists2, graph,interpolateMissing=True)
###if args.bg_exp:
### limit_bg = plot.MakeTChain(args.bg_exp, 'limit')
### best_bg = plot.TGraphFromTree(
### limit_bg, "CV", "CF", 'deltaNLL == 0')
### plot.RemoveGraphXDuplicates(best_bg)
limit = plot.MakeTChain(args.files, 'limit')
print limit
graph = plot.TGraph2DFromTree(
limit, "CV", "CF", '2*deltaNLL',
'quantileExpected > -0.5 && deltaNLL > 0 && deltaNLL < 1000')
best = plot.TGraphFromTree(limit, "CV", "CF", 'deltaNLL == 0')
plot.RemoveGraphXDuplicates(best)
hists = plot.TH2FromTGraph2D(graph, method='BinCenterAligned')
plot.fastFillTH2(hists, graph, interpolateMissing=True)
if args.bg_exp:
limit_bg = plot.MakeTChain(args.bg_exp, 'limit')
best_bg = plot.TGraphFromTree(limit_bg, "CV", "CF", 'deltaNLL == 0')
plot.RemoveGraphXDuplicates(best_bg)
# If included just plot SM exp at 1,1
if args.sm_exp:
limit_sm = plot.MakeTChain(args.sm_exp, 'limit')
best_sm = ROOT.TGraph(1, array('d', [
1,
]), array('d', [
1,
]))