def run_bonly_fit(file_name, out_file, ncpu, get_p, data_prefix="data", data_file_name="data.root", do_minos=False, cut=None):
high = 300.
rfile = R.TFile(file_name)
ws = rfile.Get("combined")
data = ws.data("obsData")
model = ws.obj("ModelConfig")
constr = model.GetNuisanceParameters()
R.RooStats.RemoveConstantParameters(constr)
# model.GetParametersOfInterest().first().setVal(0.)
# model.GetParametersOfInterest().first().setConstant()
pars = model.GetNuisanceParameters()
# run the fit
if cut:
ws.obj("obs_x_sf").setRange("fit", 10, cut)
if do_minos:
res = model.GetPdf().fitTo(data, R.RooFit.Constrain(constr), R.RooFit.Save(), R.RooFit.PrintLevel(0), R.RooFit.Minos(), R.RooFit.Hesse(), R.RooFit.Range("fit"))
else:
res = model.GetPdf().fitTo(data, R.RooFit.Constrain(constr), R.RooFit.Save(), R.RooFit.PrintLevel(0), R.RooFit.Range("fit"))
fitPars = res.floatParsFinal()
nPar = fitPars.getSize()
t = PrettyTable()
if do_minos:
t.field_names = ['#', '', "Value", "Parabolic Error", "Minos Down", "Minos Up"]
else:
t.field_names = ['#', '', "Value", "Parabolic Error"]
t.vertical_char = "&"
for i in xrange(nPar):
name = fitPars.at(i).GetName()
val = fitPars.at(i).getVal()
err = fitPars.at(i).getError()
if do_minos:
minos_up = fitPars.at(i).getErrorLo()
minos_down = fitPars.at(i).getErrorHi()
t.add_row((i, name, "{0:.3g}".format(val), "{0:.3g}".format(err), "{0:.3g}".format(minos_down), "{0:.3g}".format(minos_up)))
else:
t.add_row((i, name, "{0:.3g}".format(val), "{0:.3g}".format(err)))
print t
r = ROOT.get_results2(ws, res, cut, high)
data_results = results_to_dict(r)
f = open(out_file, 'w')
json.dump(data_results, f, indent=3)
f.close()
plot_fitted_sf(ws)
fullcor = res.correlationMatrix()
fig = plt.figure()
ax = fig.add_subplot(111)
ax.patch.set_facecolor('gray')
ax.set_aspect('equal', 'box')
# make Hinton-style correlation plot
for i in xrange(fullcor.GetNrows()):
for j in xrange(fullcor.GetNcols()):
# if i<=j: continue
c = fullcor[i][j]
if abs(c) < 0.01: continue
if c > 0: color='white'
else: color='black'
size = np.sqrt(np.abs(c))
rect = Rectangle([i-size/2, j-size/2], size, size, facecolor=color, edgecolor='black', lw=0.1)
ax.add_patch(rect)
ax.set_xlim(0, fullcor.GetNrows())
ax.set_ylim(0, fullcor.GetNrows())
plt.gca().invert_yaxis()
plt.tight_layout()
plt.savefig("plots/correlation_full_sf.pdf")
model.SetSnapshot(model.GetParametersOfInterest())
money_take2.build_background_shape(ws, 'sf', money_take2.sf_backgrounds, log=True)
money_take2.build_background_shape(ws, 'sf', money_take2.sf_backgrounds, log=False)
R.gROOT.ProcessLineSync(".L KS/AndersonDarlingTestStat.cc+")
AD = R.RooStats.AndersonDarlingTestStat(model.GetPdf())
# get the test statistic on data
ts = AD.Evaluate(data)
if get_p:
sampler = R.RooStats.ToyMCSampler(AD, 500)
sampler.SetPdf(model.GetPdf())
sampler.SetObservables(model.GetObservables())
sampler.SetGlobalObservables(model.GetGlobalObservables())
sampler.SetParametersForTestStat(model.GetParametersOfInterest())
params = R.RooArgSet()
params.add(model.GetNuisanceParameters())
params.add(model.GetParametersOfInterest())
if ncpu > 1:
pc = R.RooStats.ProofConfig(ws, ncpu, "")
sampler.SetProofConfig(pc)
sampDist = sampler.GetSamplingDistribution(params)
p = 1-sampDist.CDF(ts)
print "P value:", p
print "Test statistic on data: {:.7f}".format(ts)
plot = R.RooStats.SamplingDistPlot()
plot.AddSamplingDistribution(sampDist)
plot.Draw()
raw_input("...")
print "Test statistic on data: {:.7f}".format(ts)
return data_results
