def doLimitError(limits, unblindedStatus):
expRelErrors = []
expLabels = {}
obsRelErrors = []
obsLabels = {}
order = [0, 1, -1, 2, -2]
expErrors = [limits.expectedErrorGraph(sigma=s) for s in order]
if expErrors[0] != None:
exps = [limits.expectedGraph(sigma=s) for s in order]
expRelErrors = [(limit.divideGraph(expErrors[i], exps[i]), "ExpRelErr%d"%i) for i in xrange(len(exps))]
expLabels = {
"ExpRelErr0": "Expected median",
"ExpRelErr1": "Expected +1#sigma",
"ExpRelErr2": "Expected -1#sigma",
"ExpRelErr3": "Expected +2#sigma",
"ExpRelErr4": "Expected -2#sigma",
}
if unblindedStatus:
obsErr = limits.observedErrorGraph()
if obsErr != None:
obs = limits.observedGraph()
if obs != None:
obsRelErrors = [(limit.divideGraph(obsErr, obs), "ObsRelErr")]
obsLabels = {"ObsRelErr": "Observed"}
if len(expRelErrors) == 0 and len(obsRelErrors) == 0:
return
# Create the plot
plot = plots.PlotBase()
if len(expRelErrors) > 0:
plot.histoMgr.extendHistos([histograms.HistoGraph(x[0], x[1], drawStyle="PL", legendStyle="lp") for x in expRelErrors])
plot.histoMgr.forEachHisto(styles.generator())
def sty(h):
r = h.getRootHisto()
r.SetLineStyle(1)
r.SetLineWidth(3)
r.SetMarkerSize(1.4)
plot.histoMgr.forEachHisto(sty)
plot.histoMgr.setHistoLegendLabelMany(expLabels)
if unblindedStatus:
if len(obsRelErrors) > 0:
obsRelErrors[0][0].SetMarkerSize(1.4)
obsRelErrors[0][0].SetMarkerStyle(25)
plot.histoMgr.insertHisto(0, histograms.HistoGraph(obsRelErrors[0][0], obsRelErrors[0][1], drawStyle="PL", legendStyle="lp"))
plot.histoMgr.setHistoLegendLabelMany(obsLabels)
plot.setLegend(histograms.moveLegend(histograms.createLegend(0.48, 0.75, 0.85, 0.92), dx=0.1, dy=-0.1))
if len(limits.mass) == 1:
plot.createFrame("limitsBrRelativeUncertainty", opts={"xmin": limits.mass[0]-5.0, "xmax": limits.mass[0]+5.0, "ymin": 0, "ymaxfactor": 1.5})
else:
plot.createFrame("limitsBrRelativeUncertainty", opts={"ymin": 0, "ymaxfactor": 1.5})
plot.frame.GetXaxis().SetTitle(limit.mHplus())
plot.frame.GetYaxis().SetTitle("Uncertainty/limit")
plot.draw()
plot.setLuminosity(limits.getLuminosity())
plot.addStandardTexts()
size = 20
x = 0.2
histograms.addText(x, 0.88, limit.process, size=size)
histograms.addText(x, 0.84, limits.getFinalstateText(), size=size)
histograms.addText(x, 0.79, limit.BRassumption, size=size)
size = 22
x = 0.55
histograms.addText(x, 0.88, "Toy MC relative", size=size)
histograms.addText(x, 0.84, "statistical uncertainty", size=size)
plot.save()
def doLimitError(limits, unblindedStatus):
expRelErrors = []
expLabels = {}
obsRelErrors = []
obsLabels = {}
order = [0, 1, -1, 2, -2]
expErrors = [limits.expectedErrorGraph(sigma=s) for s in order]
if expErrors[0] != None:
exps = [limits.expectedGraph(sigma=s) for s in order]
expRelErrors = [(limit.divideGraph(expErrors[i],
exps[i]), "ExpRelErr%d" % i)
for i in xrange(len(exps))]
expLabels = {
"ExpRelErr0": "Expected median",
"ExpRelErr1": "Expected +1#sigma",
"ExpRelErr2": "Expected -1#sigma",
"ExpRelErr3": "Expected +2#sigma",
"ExpRelErr4": "Expected -2#sigma",
}
if unblindedStatus:
obsErr = limits.observedErrorGraph()
if obsErr != None:
obs = limits.observedGraph()
if obs != None:
obsRelErrors = [(limit.divideGraph(obsErr, obs), "ObsRelErr")]
obsLabels = {"ObsRelErr": "Observed"}
if len(expRelErrors) == 0 and len(obsRelErrors) == 0:
return
# Create the plot
plot = plots.PlotBase()
if len(expRelErrors) > 0:
plot.histoMgr.extendHistos([
histograms.HistoGraph(x[0], x[1], drawStyle="PL", legendStyle="lp")
for x in expRelErrors
])
plot.histoMgr.forEachHisto(styles.generator())
def sty(h):
r = h.getRootHisto()
r.SetLineStyle(1)
r.SetLineWidth(3)
r.SetMarkerSize(1.4)
plot.histoMgr.forEachHisto(sty)
plot.histoMgr.setHistoLegendLabelMany(expLabels)
if unblindedStatus:
if len(obsRelErrors) > 0:
obsRelErrors[0][0].SetMarkerSize(1.4)
obsRelErrors[0][0].SetMarkerStyle(25)
plot.histoMgr.insertHisto(
0,
histograms.HistoGraph(obsRelErrors[0][0],
obsRelErrors[0][1],
drawStyle="PL",
legendStyle="lp"))
plot.histoMgr.setHistoLegendLabelMany(obsLabels)
plot.setLegend(
histograms.moveLegend(histograms.createLegend(0.48, 0.75, 0.85, 0.92),
dx=0.1,
dy=-0.1))
if len(limits.mass) == 1:
plot.createFrame("limitsBrRelativeUncertainty",
opts={
"xmin": limits.mass[0] - 5.0,
"xmax": limits.mass[0] + 5.0,
"ymin": 0,
"ymaxfactor": 1.5
})
else:
plot.createFrame("limitsBrRelativeUncertainty",
opts={
"ymin": 0,
"ymaxfactor": 1.5
})
plot.frame.GetXaxis().SetTitle(limit.mHplus())
plot.frame.GetYaxis().SetTitle("Uncertainty/limit")
plot.draw()
plot.setLuminosity(limits.getLuminosity())
plot.addStandardTexts()
size = 20
x = 0.2
histograms.addText(x, 0.88, limit.process, size=size)
histograms.addText(x, 0.84, limits.getFinalstateText(), size=size)
histograms.addText(x, 0.79, limit.BRassumption, size=size)
size = 22
x = 0.55
histograms.addText(x, 0.88, "Toy MC relative", size=size)
histograms.addText(x, 0.84, "statistical uncertainty", size=size)
plot.save()
def doCompare(name, compareList, **kwargs):
# Define lists
legendLabels = []
limits = []
# For-loop: All label-path pairs
for label, path in compareList:
legendLabels.append(label)
dirs = glob.glob(path)
dirs.sort()
if len(dirs) == 0:
raise Exception("No directories for pattern '%s'" % path)
directory = dirs[-1]
Verbose("Picked %s" % directory, True)
limits.append(limit.BRLimits(directory, excludeMassPoints=["155"]))
# ================================================================================================================
# Do the sigma bands
# ================================================================================================================
Verbose("Creating the sigma-bands plots", True)
_opts = copy.deepcopy(opts)
doPlotSigmaBands(limits, legendLabels, opts.name + "_sigmaBands", _opts)
# ================================================================================================================
# Do expected
# ================================================================================================================
Verbose("Creating the expected plots", True)
observedList = [l.observedGraph() for l in limits]
expectedList = [l.expectedGraph() for l in limits]
# 1) Expected: Median +/- 1,2 sigma
doPlot(limits,
legendLabels,
expectedList,
opts.name + "_median",
limit.BRlimit,
_opts,
yTitle="Expected Sigma Bands")
# 2) Expected: +/- 1 sigma
list1 = [l.expectedGraph(sigma=+1) for l in limits]
list2 = [l.expectedGraph(sigma=-1) for l in limits]
exp1sigmaList = list1 + list2
legendLabels2 = legendLabels + [None] * len(legendLabels)
doPlot(limits,
legendLabels2,
exp1sigmaList,
opts.name + "_sigma1",
"Expected #pm1#sigma",
_opts,
yTitle="Expected #pm1sigma")
# 3) Expected: +/- 2 sigma
list1 = [l.expectedGraph(sigma=+2) for l in limits]
list2 = [l.expectedGraph(sigma=-2) for l in limits]
exp2sigmaList = list1 + list2
doPlot(limits,
legendLabels2,
exp2sigmaList,
opts.name + "_sigma2",
"Expected #pm2#sigma",
_opts,
yTitle="Expected #pm2sigma")
# ================================================================================================================
# Do the observed plots
# ================================================================================================================
Verbose("Creating the observed plots", True)
if opts.unblinded:
doPlot(limits,
legendLabels,
observedList,
opts.name,
limit.BRlimit,
_opts,
yTitle="Observed")
# ================================================================================================================
# Do the relative plots
# ================================================================================================================
Verbose("Creating the relative plots", True)
if not opts.relative:
return
# Overwrite some settings
_opts.yMin = 0.0
_opts.yMax = 2.5
_opts.logY = False
# 1) Relative: median
relLimits = GetRelativeLimits(limits)
expectedList = [l.expectedGraph() for l in relLimits]
doPlot(relLimits,
legendLabels,
expectedList,
opts.name + "_medianRelative",
opts.relativeYlabel,
_opts,
yTitle="Expected median")
# 2) Relative: (expected 1 sigma) / (median)
list1 = [
limit.divideGraph(l.expectedGraph(sigma=+1), l.expectedGraph())
for l in limits
]
list2 = [
limit.divideGraph(l.expectedGraph(sigma=-1), l.expectedGraph())
for l in limits
]
sigma1List = list1 + list2
doPlot(limits,
legendLabels2,
sigma1List,
opts.name + "_sigma1Relative",
"Expected #pm1#sigma / median",
_opts,
yTitle="Expected #pm1#sigma / median")
# 3) Relative: (expected 2 sigma) / (median)
list1 = [
limit.divideGraph(l.expectedGraph(sigma=+2), l.expectedGraph())
for l in limits
]
list2 = [
limit.divideGraph(l.expectedGraph(sigma=-2), l.expectedGraph())
for l in limits
]
sigma2List = list1 + list2
doPlot(limits,
legendLabels2,
sigma2List,
opts.name + "_sigma2Relative",
"Expected #pm2#sigma / median",
_opts,
yTitle="Expected #pm2#sigma / median")
return
def doCompare(name, compareList, opts, **kwargs):
legendLabels = []
limits = []
for label, path in compareList:
legendLabels.append(label)
dirs = glob.glob(path)
dirs.sort()
if len(dirs) == 0:
raise Exception("No directories for pattern '%s'" % path)
directory = dirs[-1]
print "Picked %s" % directory
limits.append(limit.BRLimits(directory, excludeMassPoints=["155"]))
doPlot2(limits, legendLabels, name)
limitOpts = kwargs.get(
"limitOpts",
{"ymax": _ifNotNone(opts.ymax, limits[0].getFinalstateYmaxBR())})
expectedSigmaRelativeOpts = kwargs.get("expectedSigmaRelativeOpts",
{"ymaxfactor": 1.2})
moveLegend = kwargs.get("moveLegend", {})
doPlot(limits,
legendLabels, [l.observedGraph() for l in limits],
name + "_observed",
limit.BRlimit,
opts=kwargs.get("observedOpts", limitOpts),
moveLegend=moveLegend,
plotLabel="Observed")
doPlot(limits,
legendLabels, [l.expectedGraph() for l in limits],
name + "_expectedMedian",
limit.BRlimit,
opts=kwargs.get("expectedMedianOpts", limitOpts),
moveLegend=moveLegend,
log=kwargs.get("log", False),
plotLabel="Expected median")
if opts.relative:
for i in range(1, len(limits)):
limits[i].divideByLimit(limits[0])
# Set reference to 1
for j in range(0, len(limits[0].expectedMedian)):
limits[0].expectedMedian[j] = 1.0
limits[0].expectedMinus2[j] = 1.0
limits[0].expectedMinus1[j] = 1.0
limits[0].expectedPlus2[j] = 1.0
limits[0].expectedPlus1[j] = 1.0
limits[0].observed[j] = 1.0
# Set y scale and require it to be linear
kwargs["expectedMedianOptsRelative"] = {
"ymin": 0.5,
"ymax": _ifNotNone(opts.relativeYmax, 1.5)
}
kwargs["log"] = False
doPlot(limits,
legendLabels, [l.expectedGraph() for l in limits],
name + "_expectedMedianRelative",
opts.relativeYlabel,
opts=kwargs.get("expectedMedianOptsRelative", limitOpts),
moveLegend=moveLegend,
log=kwargs.get("log", False),
plotLabel="Expected median")
print "Skipping +-1 and 2 sigma plots for --relative"
sys.exit()
if opts.relativePairs:
if len(limits) % 2 != 0:
print "Number of limits is not even!"
sys.exit(1)
divPoint = len(limits) / 2
denoms = limits[:divPoint]
numers = limits[divPoint:]
for i in xrange(0, divPoint):
numers[i].divideByLimit(denoms[i])
doPlot(numers,
legendLabels[:divPoint], [l.expectedGraph() for l in numers],
name + "_expectedMedianRelative",
opts.relativeYlabel,
opts={
"ymin": 0.5,
"ymax": _ifNotNone(opts.relativeYmax, 1.5)
},
plotLabel="Expected median")
print "Skipping +-1 and 2 sigma plots for --relativePairs"
sys.exit()
legendLabels2 = legendLabels + [None] * len(legendLabels)
doPlot(limits,
legendLabels2,
[
limit.divideGraph(l.expectedGraph(sigma=+1), l.expectedGraph())
for l in limits
] + [
limit.divideGraph(l.expectedGraph(sigma=-1), l.expectedGraph())
for l in limits
],
name + "_expectedSigma1Relative",
"Expected #pm1#sigma / median",
opts=kwargs.get("expectedSigma1RelativeOpts",
expectedSigmaRelativeOpts),
moveLegend=moveLegend,
plotLabel="Expected #pm1#sigma / median")
doPlot(limits,
legendLabels2,
[
limit.divideGraph(l.expectedGraph(sigma=+2), l.expectedGraph())
for l in limits
] + [
limit.divideGraph(l.expectedGraph(sigma=-2), l.expectedGraph())
for l in limits
],
name + "_expectedSigma2Relative",
"Expected #pm2#sigma / median",
opts=kwargs.get("expectedSigma2RelativeOpts",
expectedSigmaRelativeOpts),
moveLegend=moveLegend,
plotLabel="Expected #pm2#sigma / median")
doPlot(limits,
legendLabels2,
[l.expectedGraph(sigma=+1)
for l in limits] + [l.expectedGraph(sigma=-1) for l in limits],
name + "_expectedSigma1",
"Expected #pm1#sigma",
opts=kwargs.get("expectedSigma1Opts", limitOpts),
moveLegend=moveLegend,
plotLabel="Expexted #pm1sigma")
doPlot(limits,
legendLabels2,
[l.expectedGraph(sigma=+2)
for l in limits] + [l.expectedGraph(sigma=-2) for l in limits],
name + "_expectedSigma2",
"Expected #pm2#sigma",
opts=kwargs.get("expectedSigma2Opts", limitOpts),
moveLegend=moveLegend,
plotLabel="Expected #pm2sigma")
def doCompare(name, compareList, opts, **kwargs):
legendLabels = []
limits = []
for label, path in compareList:
legendLabels.append(label)
dirs = glob.glob(path)
dirs.sort()
if len(dirs) == 0:
raise Exception("No directories for pattern '%s'" % path)
directory = dirs[-1]
print "Picked %s" % directory
limits.append(limit.BRLimits(directory, excludeMassPoints=["155"]))
doPlot2(limits, legendLabels, name)
limitOpts = kwargs.get("limitOpts", {"ymax": _ifNotNone(opts.ymax, limits[0].getFinalstateYmaxBR())})
expectedSigmaRelativeOpts = kwargs.get("expectedSigmaRelativeOpts", {"ymaxfactor": 1.2})
moveLegend = kwargs.get("moveLegend", {})
doPlot(limits, legendLabels, [l.observedGraph() for l in limits],
name+"_observed", limit.BRlimit, opts=kwargs.get("observedOpts", limitOpts), moveLegend=moveLegend,
plotLabel="Observed")
doPlot(limits, legendLabels, [l.expectedGraph() for l in limits],
name+"_expectedMedian", limit.BRlimit, opts=kwargs.get("expectedMedianOpts", limitOpts), moveLegend=moveLegend, log=kwargs.get("log",False),
plotLabel="Expected median")
if opts.relative:
for i in range(1, len(limits)):
limits[i].divideByLimit(limits[0])
# Set reference to 1
for j in range(0, len(limits[0].expectedMedian)):
limits[0].expectedMedian[j] = 1.0
limits[0].expectedMinus2[j] = 1.0
limits[0].expectedMinus1[j] = 1.0
limits[0].expectedPlus2[j] = 1.0
limits[0].expectedPlus1[j] = 1.0
limits[0].observed[j] = 1.0
# Set y scale and require it to be linear
kwargs["expectedMedianOptsRelative"] = {"ymin": 0.5, "ymax": _ifNotNone(opts.relativeYmax, 1.5)}
kwargs["log"] = False
doPlot(limits, legendLabels, [l.expectedGraph() for l in limits],
name+"_expectedMedianRelative", opts.relativeYlabel, opts=kwargs.get("expectedMedianOptsRelative", limitOpts), moveLegend=moveLegend, log=kwargs.get("log",False),
plotLabel="Expected median")
print "Skipping +-1 and 2 sigma plots for --relative"
sys.exit()
if opts.relativePairs:
if len(limits) % 2 != 0:
print "Number of limits is not even!"
sys.exit(1)
divPoint = len(limits) / 2
denoms = limits[:divPoint]
numers = limits[divPoint:]
for i in xrange(0, divPoint):
numers[i].divideByLimit(denoms[i])
doPlot(numers, legendLabels[:divPoint], [l.expectedGraph() for l in numers],
name+"_expectedMedianRelative", opts.relativeYlabel, opts={"ymin": 0.5, "ymax": _ifNotNone(opts.relativeYmax, 1.5)},
plotLabel="Expected median")
print "Skipping +-1 and 2 sigma plots for --relativePairs"
sys.exit()
legendLabels2 = legendLabels + [None]*len(legendLabels)
doPlot(limits, legendLabels2,
[limit.divideGraph(l.expectedGraph(sigma=+1), l.expectedGraph()) for l in limits] +
[limit.divideGraph(l.expectedGraph(sigma=-1), l.expectedGraph()) for l in limits],
name+"_expectedSigma1Relative", "Expected #pm1#sigma / median", opts=kwargs.get("expectedSigma1RelativeOpts", expectedSigmaRelativeOpts), moveLegend=moveLegend,
plotLabel="Expected #pm1#sigma / median")
doPlot(limits, legendLabels2,
[limit.divideGraph(l.expectedGraph(sigma=+2), l.expectedGraph()) for l in limits] +
[limit.divideGraph(l.expectedGraph(sigma=-2), l.expectedGraph()) for l in limits],
name+"_expectedSigma2Relative", "Expected #pm2#sigma / median", opts=kwargs.get("expectedSigma2RelativeOpts", expectedSigmaRelativeOpts), moveLegend=moveLegend,
plotLabel="Expected #pm2#sigma / median")
doPlot(limits, legendLabels2,
[l.expectedGraph(sigma=+1) for l in limits] +
[l.expectedGraph(sigma=-1) for l in limits],
name+"_expectedSigma1", "Expected #pm1#sigma", opts=kwargs.get("expectedSigma1Opts", limitOpts), moveLegend=moveLegend,
plotLabel="Expexted #pm1sigma")
doPlot(limits, legendLabels2,
[l.expectedGraph(sigma=+2) for l in limits] +
[l.expectedGraph(sigma=-2) for l in limits],
name+"_expectedSigma2", "Expected #pm2#sigma", opts=kwargs.get("expectedSigma2Opts", limitOpts), moveLegend=moveLegend,
plotLabel="Expected #pm2sigma")
def doCompare(name, compareList, gOpts, **kwargs):
# Define lists
legendLabels = []
limits = []
# For-loop: All label-path pairs
for label, path in compareList:
legendLabels.append(label)
dirs = glob.glob(path)
dirs.sort()
if len(dirs) == 0:
raise Exception("No directories for pattern '%s'" % path)
directory = dirs[-1]
Print("Picked %s" % directory, True)
limits.append(limit.BRLimits(directory, excludeMassPoints=["155"]))
# Do the plot
doPlotSigmaBands(limits, legendLabels, gOpts.name)
# Define the graph lists
observedList = [l.observedGraph() for l in limits]
expectedList = [l.expectedGraph() for l in limits]
doPlot(limits,
legendLabels,
expectedList,
gOpts.name + "_expectedMedian",
limit.BRlimit,
myOpts=gOpts,
plotLabel="Expected median")
if gOpts.unblinded:
doPlot(limits,
legendLabels,
observedList,
gOpts.name + "_observed",
limit.BRlimit,
myOpts=gOpts,
plotLabel="Observed")
if gOpts.relative:
nLimits = len(limits)
# For-loop: All limits
for i in range(1, nLimits):
limits[i].divideByLimit(limits[0])
# Set reference values to 1
for j in range(0, len(limits[0].expectedMedian)):
limits[0].expectedMedian[j] = 1.0
limits[0].expectedMinus2[j] = 1.0
limits[0].expectedMinus1[j] = 1.0
limits[0].expectedPlus2[j] = 1.0
limits[0].expectedPlus1[j] = 1.0
limits[0].observed[j] = 1.0
# Do the relative plot
doPlot(limits,
legendLabels,
expectedList,
gOpts.name + "_expectedMedianRelative",
gOpts.relativeYlabel,
myOpts=gOpts,
plotLabel="Expected median")
Print("Skipping +-1 and 2 sigma plots for --relative", True)
sys.exit()
if gOpts.relativePairs:
if len(limits) % 2 != 0:
Print("Number of limits is not even!", True)
sys.exit(1)
divPoint = len(limits) / 2
denoms = limits[:divPoint]
numers = limits[divPoint:]
# For-loop: All division points
for i in xrange(0, divPoint):
numers[i].divideByLimit(denoms[i])
expectedNumersList = [l.expectedGraph() for l in numers]
doPlot(numers,
legendLabels[:divPoint],
gexpectedNumersList,
gOpts.name + "_expectedMedianRelative",
gOpts.relativeYlabel,
plotLabel="Expected median")
Print("Skipping +-1 and 2 sigma plots for --relativePairs", True)
sys.exit()
legendLabels2 = legendLabels + [None] * len(legendLabels)
doPlot(limits,
legendLabels2,
[
limit.divideGraph(l.expectedGraph(sigma=+1), l.expectedGraph())
for l in limits
] + [
limit.divideGraph(l.expectedGraph(sigma=-1), l.expectedGraph())
for l in limits
],
gOpts.name + "_expectedSigma1Relative",
"Expected #pm1#sigma / median",
myOpts=gOpts,
plotLabel="Expected #pm1#sigma / median")
doPlot(limits,
legendLabels2,
[
limit.divideGraph(l.expectedGraph(sigma=+2), l.expectedGraph())
for l in limits
] + [
limit.divideGraph(l.expectedGraph(sigma=-2), l.expectedGraph())
for l in limits
],
gOpts.name + "_expectedSigma2Relative",
"Expected #pm2#sigma / median",
myOpts=gOpts,
plotLabel="Expected #pm2#sigma / median")
doPlot(limits,
legendLabels2,
[l.expectedGraph(sigma=+1)
for l in limits] + [l.expectedGraph(sigma=-1) for l in limits],
gOpts.name + "_expectedSigma1",
"Expected #pm1#sigma",
myOpts=gOpts,
plotLabel="Expexted #pm1sigma")
doPlot(limits,
legendLabels2,
[l.expectedGraph(sigma=+2)
for l in limits] + [l.expectedGraph(sigma=-2) for l in limits],
gOpts.name + "_expectedSigma2",
"Expected #pm2#sigma",
myOpts=gOpts,
plotLabel="Expected #pm2sigma")
return
