def process_one_file(infn):
# load precalculated matrices
tinf = TFile(infn)
tcov = tinf.Get('cov_zero_sup')
cov = matrix(tcov)
tinvcov = tinf.Get('covinv_zero_sup')
invcov = matrix(tinvcov)
# calculate the global correlation coefficient for each bin
rhos = []
mat_size = cov.shape[0]
for i in range(mat_size):
mult = cov[i, i] * invcov[i, i]
# if mult < 0:
# print('*****WARNING*****\nnegative number at {} with value {}'.format(i, mult))
if mult >= 1:
rhos.append(math.sqrt(1. - 1. / mult))
# debugging info
print('size of the rho vector for {} iterations:\n{}'.format(
find_niter(infn), len(rhos)))
# return results
if len(rhos) == 0: return 0, 0
return statistics.mean(rhos), len(rhos)
def test_matrix():
for cls in (getattr(ROOT, 'TMatrix{0}'.format(atype)) for atype in 'DF'):
m = cls(5, 5)
m[1][2] = 2
n = rnp.matrix(m)
assert_equal(n[1, 2], 2)
for cls in (getattr(ROOT, 'TMatrix{0}Sym'.format(atype)) for atype in 'DF'):
m = cls(5)
m[2][2] = 2
n = rnp.matrix(m)
assert_equal(n[2, 2], 2)
assert_raises(TypeError, rnp.matrix, object)
def preprocess_covmat(iter, tolerance):
tinf = TFile('input_files/rooUnfold{}iter.root'.format(iter))
covmat = tinf.Get('cv/EAvail/Covariance')
cov = matrix(covmat)
# zero row/column supression
# supress columns and rows with all elements zero
nonzero_row_indices = [
i for i in range(cov.shape[0]) if not np.allclose(cov[i, :], 0)
]
nonzero_col_indices = [
i for i in range(cov.shape[1]) if not np.allclose(cov[:, i], 0)
]
cov_sup = cov[nonzero_row_indices, :]
cov_sup = cov_sup[:, nonzero_col_indices]
invcov = np.linalg.pinv(cov_sup, rcond=tolerance)
return cov_sup, invcov
def check_matrix_sym(cls):
mat = cls(5)
mat[2][2] = 2
np_mat = rnp.matrix(mat)
assert_equal(np_mat[2, 2], 2)
def check_matrix(cls):
mat = cls(5, 5)
mat[1][2] = 2
np_mat = rnp.matrix(mat)
assert_equal(np_mat[1, 2], 2)
def bkg_unc_propagation(self, outputfile):
# define the set obs = (x)
self.wspace.defineSet('obs', 'x')
# make the set obs known to Python
obs = self.wspace.set('obs')
nsig_interested_pdf = self.pdf['sig'].createIntegral(
obs, obs, "fom_window")
nsig_interested_pdf_err = nsig_interested_pdf.getPropagatedError(
self.results, obs)
self.nsig_interested = self.var['nsig'].getVal(
) * nsig_interested_pdf.getVal()
self.nsig_interested_err = self.nsig_interested * np.sqrt(
pow(self.var['nsig'].getError() / self.var['nsig'].getVal(), 2) +
pow(nsig_interested_pdf_err / nsig_interested_pdf.getVal(), 2)
) if self.var['nsig'].getVal() != 0.0 else 0.0
nbkg_comb_pdf = self.pdf['bkg'].createIntegral(obs, obs, "fom_window")
nbkg_comb_pdf_err = nbkg_comb_pdf.getPropagatedError(self.results, obs)
nbkg_comb = self.var['nbkg'].getVal() * nbkg_comb_pdf.getVal()
nbkg_comb_err = nbkg_comb * np.sqrt(
pow(self.var['nbkg'].getError() / self.var['nbkg'].getVal(), 2) +
pow(nbkg_comb_pdf_err / nbkg_comb_pdf.getVal(), 2)
) if self.var['nbkg'].getVal() != 0.0 else 0.0
self.nbkg_total = nbkg_comb
print("*" * 80)
print("MVA Cut: {}".format(self.mvaCut))
if not self.fitConverged:
print("*" * 20 + "NOT COVERGE" + "*" * 20)
print("Number of signals: {}".format(self.var['nsig'].getVal()))
print("Number of signals in 3.0 sigma: {}, uncertainty: {}".format(
self.nsig_interested, self.nsig_interested_err))
print(
"Number of background - combinatorial: {}, uncertainty: {}".format(
nbkg_comb, nbkg_comb_err))
for name in partialfit.keys():
nbkg_pdf_pdf = self.pdf[name].createIntegral(
obs, obs, "fom_window")
nbkg_partial = self.var[name].getVal() * nbkg_pdf_pdf.getVal()
self.nbkg_total += nbkg_partial
print("Number of background - {}: {}".format(name, nbkg_partial))
# Calculate 1-sigma error band of the total bkg through linear error propagation
bkgframe = self.mass.frame()
self.data.plotOn(bkgframe, RooFit.Binning(self.nbin_data))
nbinx = 1000
xvar = np.linspace(self.fom_low, self.fom_up, nbinx)
self.fit_params = self.pdf['model'].getVariables()
ordered_fit_params = ['exp_alpha', 'nbkg'] + [
'n' + name for name in self.partialfit.keys()
]
if not self.blinded:
ordered_fit_params += [
'nsig',
]
full_bkg = [
'bkg',
] + [name for name in self.partialfit.keys()]
fit_params_info = OrderedDict()
for name in ordered_fit_params:
fit_params_info[name] = {
'mean': self.fit_params.find(name).getVal(),
'error': self.fit_params.find(name).getError()
}
self.pdf['model'].plotOn(bkgframe,
RooFit.Components(",".join(full_bkg)))
model_curve = bkgframe.getCurve()
model_cen = np.array([model_curve.interpolate(x) for x in xvar])
bkgframe.remove(str(0), False)
#self.results.covarianceMatrix().Print()
#self.results.correlationMatrix().Print()
covMatrix = root_numpy.matrix(self.results.covarianceMatrix())
exp_event = self.pdf['model'].expectedEvents(self.fit_params)
fa = []
for name, info in fit_params_info.items():
adjust_norm = info['error'] if (name in ([
'nsig',
'nbkg',
] + ['n' + p for p in self.partialfit.keys()])) else 0.0
self.fit_params.setRealValue(name, info['mean'] + info['error'])
self.pdf['model'].plotOn(
bkgframe, RooFit.Components(",".join(full_bkg)),
RooFit.Normalization(exp_event + adjust_norm,
ROOT.RooAbsReal.NumEvent))
model_curve = bkgframe.getCurve()
fa_plus = np.array([model_curve.interpolate(x) for x in xvar])
bkgframe.remove(str(0), False)
self.fit_params.setRealValue(name,
info['mean'] - 2.0 * info['error'])
self.pdf['model'].plotOn(
bkgframe, RooFit.Components(",".join(full_bkg)),
RooFit.Normalization(exp_event - adjust_norm,
ROOT.RooAbsReal.NumEvent))
model_curve = bkgframe.getCurve()
fa_minus = np.array([model_curve.interpolate(x) for x in xvar])
bkgframe.remove(str(0), False)
if name == 'nsig':
fa.append(np.zeros(nbinx))
else:
fa.append((fa_plus - fa_minus) / (2.0 * info['error']))
# reset the params matrix
self.fit_params.setRealValue(name, info['mean'])
fa = np.array(fa).T
tmp = np.array(
[np.asarray(np.matmul(FA, covMatrix)).flatten() for FA in fa])
bkg_unc = np.sqrt(np.array([np.dot(t, FA) for t, FA in zip(tmp, fa)]))
self.nbkg_total_err = np.sqrt(np.trapz(bkg_unc * bkg_unc, x=xvar)) / (
(self.fit_up - self.fit_low) / self.nbin_data)
if self.plotBkgUnc:
fig, ax = plt.subplots()
ax.plot(xvar,
model_cen,
'b-',
label=r'$N_{{\rm bkg}}={0:.1f}\pm{1:.1f}$'.format(
self.nbkg_total, self.nbkg_total_err))
ax.fill_between(xvar,
model_cen - bkg_unc,
model_cen + bkg_unc,
facecolor='red',
alpha=0.5,
linewidth=0.0,
label=r'$1\sigma$')
ax.set_xlabel(r'$m(K^{+}e^{+}e^{-}) [{\rm GeV}]$')
ax.set_ylabel(r'a.u.')
ax.set_ylim(bottom=0)
ax.legend(loc='upper right')
fig.savefig(outputfile.replace('.pdf', '') +
'_totalbkg_1sigma.pdf',
bbox_inches='tight')
self.SNR = self.nsig_interested / np.sqrt(self.nsig_interested +
self.nbkg_total)
print("Total number of background: {}, uncertainty: {}".format(
self.nbkg_total, self.nbkg_total_err))
print("S/sqrt(S+B): {}".format(self.SNR))
print("*" * 80)
def check_matrix_sym(cls):
mat = cls(5)
mat[2][2] = 2
np_mat = rnp.matrix(mat)
assert_equal(np_mat[2, 2], 2)
def check_matrix(cls):
mat = cls(5, 5)
mat[1][2] = 2
np_mat = rnp.matrix(mat)
assert_equal(np_mat[1, 2], 2)
'--tolerance',
type=str,
default='1e-10',
help='Tolerance for TSVD.')
args = parser.parse_args()
tolerance = float(args.tolerance)
# grab the input file with the covariance matrix to play with
inpn = '../grid_output/unfolded_spectra_roounfold_iter_{}_zero_suppress.root'.format(
args.iterations)
intf = TFile(inpn)
errmat = intf.Get('cov_zero_sup')
herrmat = intf.Get('hcov_zero_sup')
# operate in python
covmat = matrix(errmat)
covinv = np.linalg.pinv(covmat, rcond=tolerance)
# transform back to ROOT
tcovinv = make_tmatrix(covinv)
hcovinv = TH2F(tcovinv)
# save results
dest_folder = 'stored_matrices/{}'.format(tolerance)
if not os.path.exists(dest_folder):
os.makedirs(dest_folder)
toutf = TFile(
os.path.join(dest_folder, 'iter{}.root'.format(args.iterations)),
'recreate')
errmat.Write('cov_zero_sup')
herrmat.Write('hcov_zero_sup')
def fit(tree, outputfile, **kwargs):
sigPDF = kwargs.get('sigPDF', 0)
bkgPDF = kwargs.get('bkgPDF', 0)
isMC = kwargs.get('isMC', False)
drawSNR = kwargs.get('drawSNR', False)
mvaCut = kwargs.get('mvaCut', 0.0)
blinded = kwargs.get('blinded', False)
sig_low = kwargs.get('sig_low', BLIND_LOW)
sig_up = kwargs.get('sig_up', BLIND_UP)
fom_low = kwargs.get('fom_low', B_FOM_LOW)
fom_up = kwargs.get('fom_up', B_FOM_UP)
expS = kwargs.get('expS', 0)
params = kwargs.get('params', {})
sigName = kwargs.get('sigName', "B^{+}#rightarrow K^{+} J/#psi(#rightarrow e^{+}e^{-})")
floatSig = kwargs.get('floatSig', False)
prefix = kwargs.get('prefix', 'BToKEE')
partialfit = kwargs.get('partialfit', {})
fitmode = kwargs.get('fitmode', 'b')
plotSigmaBand = kwargs.get('plotSigmaBand', False)
msgservice = ROOT.RooMsgService.instance()
msgservice.setGlobalKillBelow(RooFit.FATAL)
wspace = ROOT.RooWorkspace('myWorkSpace')
ROOT.gStyle.SetOptFit(0000);
ROOT.gROOT.SetBatch(True);
ROOT.gROOT.SetStyle("Plain");
ROOT.gStyle.SetGridStyle(3);
ROOT.gStyle.SetOptStat(000000);
ROOT.gStyle.SetOptTitle(0)
thevars = ROOT.RooArgSet()
if fitmode == 'jpsi':
xmin, xmax = 2.5, 3.5
#xmin, xmax = 3.0, 4.0
bMass = ROOT.RooRealVar("{}_mll_fullfit".format(prefix), "m(e^{+}e^{-})", 2.0, 5.0, "GeV")
wspace.factory('mean[3.096916, 2.9, 3.8]')
elif fitmode == 'phi':
xmin, xmax = 0.98, 1.06
bMass = ROOT.RooRealVar("{}_fit_phi_mass".format(prefix), "m(K^{+}K^{-})", 0.98, 1.06, "GeV")
wspace.factory('mean[1.02, 0.98, 1.06]')
else:
bMass = ROOT.RooRealVar("{}_fit_mass".format(prefix), "m(K^{+}e^{+}e^{-})", 4.0, 6.0, "GeV")
xmin, xmax = FIT_LOW, FIT_UP
#xmin, xmax = FIT_LOW, 5.7
wspace.factory('mean[5.272e+00, 5.22e+00, 5.5e+00]')
thevars.add(bMass)
fulldata = ROOT.RooDataSet('fulldata', 'fulldata', tree, ROOT.RooArgSet(thevars))
theBMassfunc = ROOT.RooFormulaVar("x", "x", "@0", ROOT.RooArgList(bMass) )
theBMass = fulldata.addColumn(theBMassfunc) ;
theBMass.setRange(xmin,xmax);
thevars.add(theBMass)
cut = ''
#print cut
data = fulldata.reduce(thevars, cut)
getattr(wspace,'import')(data, RooFit.Rename("data"))
wspace.factory('nsig[5000.0, 0.0, 1000000.0]' )
wspace.factory('nbkg[10000.0, 0.0, 1000000.0]')
if sigPDF == 0:
# Double-sided Crystal-ball
wspace.factory('width[4.1858e-02, 1.0e-6, 5.0e-1]')
wspace.factory('alpha1[1.0, 0.0, 10.0]')
wspace.factory('n1[1.0, 1.0, 20.0]')
wspace.factory('alpha2[1.0, 0.0, 10.0]')
wspace.factory('n2[1.0, 1.0, 20.0]')
wspace.factory('GenericPdf::sig("DoubleCBFast(x,mean,width,alpha1,n1,alpha2,n2)", {x,mean,width,alpha1,n1,alpha2,n2})')
mean = wspace.var('mean')
width = wspace.var('width')
alpha1 = wspace.var('alpha1')
n1 = wspace.var('n1')
alpha2 = wspace.var('alpha2')
n2 = wspace.var('n2')
if sigPDF == 1:
# Voigtian
wspace.factory('width[1.000e-02, 1.000e-04, 1.000e-01]')
wspace.factory('sigma[7.1858e-02, 1.e-3, 1.e-1]')
wspace.factory('Voigtian::sig(x,mean,width,sigma)')
if sigPDF == 2:
# Gaussian
wspace.factory('sigma[7.1858e-02, 1.0e-3, 5.0e-1]')
wspace.factory('Gaussian::sig(x,mean,sigma)')
if sigPDF == 3:
# Crystal-ball
wspace.factory('sigma[7.1858e-02, 1.0e-6, 5.0e-1]')
wspace.factory('alpha[1.0, 0.0, 10.0]')
wspace.factory('n[2, 1, 10]')
wspace.factory('CBShape::sig(x,mean,sigma,alpha,n)')
if sigPDF == 4:
# Two Double-sided Crystal-ball
wspace.factory('width[7.1858e-02, 1.0e-6, 5.0e-1]')
wspace.factory('alpha1[1.0, 0.0, 10.0]')
wspace.factory('n1[2.0, 1.0, 20.0]')
wspace.factory('alpha2[1.0, 0.0, 10.0]')
wspace.factory('n2[2.0, 1.0, 20.0]')
wspace.factory('GenericPdf::cb("DoubleCBFast(x,mean,width,alpha1,n1,alpha2,n2)", {x,mean,width,alpha1,n1,alpha2,n2})')
wspace.factory('sigma[7.1858e-03, 1.0e-6, 5.0e-1]')
wspace.factory('Gaussian::gaus(x,mean,sigma)')
wspace.factory('f1[0.5, 0.0, 1.0]')
wspace.factory('SUM::sig(f1*cb, gaus)')
if bkgPDF == 0:
# Exponential
wspace.factory('exp_alpha[-1.0, -100.0, -1.e-4]')
exp_alpha = wspace.var('exp_alpha')
wspace.factory('Exponential::bkg(x,exp_alpha)')
if bkgPDF == 1:
# Polynomial
wspace.factory('c0[1.0, -1.0, 1.0]')
wspace.factory('c1[-0.1, -1.0, 1.0]')
wspace.factory('c2[-0.1, -1.0, 1.0]')
wspace.factory('Chebychev::bkg(x,{c0,c1,c2})')
if bkgPDF == 2:
wspace.factory('c1[0.0, -100.0, 100.0]')
wspace.factory('Polynomial::bkg(x,{c1})')
if isMC:
wspace.factory('ExtendPdf::model(sig,nsig)')
else:
npartial = {}
for name, info in partialfit.items():
wpf = ROOT.TFile(info['filename'], "READ")
wp = wpf.Get("myPartialWorkSpace")
partialPDF = wp.pdf(name)
wspace.factory('n{}[{}, {}, {}]'.format(name, info['expected_yield'], info['expected_yield'] - 4.0*np.sqrt(info['expected_yield']), info['expected_yield'] + 4.0*np.sqrt(info['expected_yield'])) if 'expected_yield' in info else 'n{}[10.0, 0.0, 100000.0]'.format(name))
getattr(wspace, "import")(partialPDF, RooFit.Rename(name))
npartial[name] = wspace.var('n'+name)
#if 'expected_yield' in info:
#npartial[name].setVal(info['expected_yield'])
#npartial[name].setConstant(True)
wspace.factory('SUM::model(nsig*sig,nbkg*bkg{})'.format(','+','.join(['n'+name+'*'+name for name in partialfit.keys()])))
if not floatSig:
mean.setVal(params['mean'])
width.setVal(params['width'])
alpha1.setVal(params['alpha1'])
n1.setVal(params['n1'])
alpha2.setVal(params['alpha2'])
n2.setVal(params['n2'])
mean.setConstant(True)
width.setConstant(True)
alpha1.setConstant(True)
n1.setConstant(True)
alpha2.setConstant(True)
n2.setConstant(True)
model = wspace.pdf('model')
bkg = wspace.pdf('bkg')
sig = wspace.pdf('sig')
nsig = wspace.var('nsig')
if blinded:
nsig.setVal(expS)
#nsig.setVal(0.0)
nsig.setConstant(True)
nbkg = wspace.var('nbkg')
partial_pdf = {}
for name in partialfit.keys():
partial_pdf[name] = wspace.pdf(name)
# define the set obs = (x)
wspace.defineSet('obs', 'x')
# make the set obs known to Python
obs = wspace.set('obs')
theBMass.setRange("window",sig_low,sig_up)
theBMass.setRange("fom_window",fom_low,fom_up)
theBMass.setRange("SB1",xmin,sig_low)
theBMass.setRange("SB2",sig_up,xmax)
## fit the model to the data.
print('Fitting data...')
if blinded:
results = model.fitTo(data, RooFit.Extended(True), RooFit.Save(), RooFit.Range("SB1,SB2"), RooFit.SplitRange(True), RooFit.PrintLevel(-1))
#nsig.setVal(expS)
else:
results = model.fitTo(data, RooFit.Extended(True), RooFit.Save(), RooFit.Range(xmin,xmax), RooFit.PrintLevel(-1))
results.Print()
fitConverged = True if results.status() == 0 else False
#fitConverged = False
nbin_data = 50
if isMC:
fracSigRange = sig.createIntegral(obs,obs,"fom_window") ;
print(data.sumEntries(),fracSigRange.getVal())
else:
nsig_interested_pdf = sig.createIntegral(obs,obs,"fom_window") ;
nsig_interested_pdf_err = nsig_interested_pdf.getPropagatedError(results, obs)
nsig_interested = nsig.getVal() * nsig_interested_pdf.getVal()
nsig_interested_err = nsig_interested * np.sqrt(pow(nsig.getError()/nsig.getVal(), 2) + pow(nsig_interested_pdf_err/nsig_interested_pdf.getVal(), 2)) if nsig.getVal() != 0.0 else 0.0
nbkg_comb_pdf = bkg.createIntegral(obs,obs,"fom_window")
nbkg_comb_pdf_err = nbkg_comb_pdf.getPropagatedError(results, obs)
nbkg_comb = nbkg.getVal() * nbkg_comb_pdf.getVal()
nbkg_comb_err = nbkg_comb * np.sqrt(pow(nbkg.getError()/nbkg.getVal(), 2) + pow(nbkg_comb_pdf_err/nbkg_comb_pdf.getVal(), 2)) if nbkg.getVal() != 0.0 else 0.0
nbkg_total = nbkg_comb
print("*"*80)
print("MVA Cut: {}".format(mvaCut))
if not fitConverged:
print("*"*20 + "NOT COVERGE" + "*"*20)
print("Number of signals: {}".format(nsig.getVal()))
print("Number of signals in 1.4 sigma: {}, uncertainty: {}".format(nsig_interested, nsig_interested_err))
print("Number of background - combinatorial: {}, uncertainty: {}".format(nbkg_comb, nbkg_comb_err))
for name, nvar in npartial.items():
nbkg_pdf_pdf = partial_pdf[name].createIntegral(obs,obs,"fom_window")
nbkg_partial = nvar.getVal() * nbkg_pdf_pdf.getVal()
nbkg_total += nbkg_partial
print("Number of background - {}: {}".format(name, nbkg_partial))
# Calculate 1-sigma error band of the total bkg through linear error propagation
#xvar = np.linspace(xmin, xmax, nbin_data, endpoint=False) + ( (xmax - xmin) / (nbin_data * 2.0) )
#xvar = np.array([x for x in xvar if ((x > sig_low) and x < sig_up)])
#nbinx = xvar.shape[0]
#bkgframe = ROOT.RooPlot()
bkgframe = theBMass.frame()
#bkgframe = xframe
#bkgframe = xframe.emptyClone('test')
data.plotOn(bkgframe, RooFit.Binning(nbin_data), RooFit.Name("data"))
nd = 1.0
nbinx = 1000
xvar = np.linspace(fom_low, fom_up, nbinx)
fit_params = model.getVariables()
ordered_fit_params = ['exp_alpha', 'nbkg'] + ['n'+name for name in partialfit.keys()]
if not blinded:
ordered_fit_params += ['nsig',]
full_bkg = ['bkg',] + [name for name in partialfit.keys()]
#full_bkg = ['bkg',]
fit_params_info = OrderedDict()
for name in ordered_fit_params:
fit_params_info[name] = {'mean': fit_params.find(name).getVal(), 'error': fit_params.find(name).getError()}
model.plotOn(bkgframe,RooFit.Components(",".join(full_bkg)),RooFit.Normalization(nd, ROOT.RooAbsReal.Relative))
model_curve = bkgframe.getCurve()
model_cen = np.array([model_curve.interpolate(x) for x in xvar])
#print(model_cen)
bkgframe.remove(str(0),False)
#results.covarianceMatrix().Print()
#results.correlationMatrix().Print()
covMatrix = root_numpy.matrix(results.covarianceMatrix())
exp_event = model.expectedEvents(fit_params)
fa = []
for name, info in fit_params_info.items():
adjust_norm = info['error'] if (name in (['nsig', 'nbkg',] + ['n'+p for p in partialfit.keys()])) else 0.0
fit_params.setRealValue(name, info['mean']+info['error'])
model.plotOn(bkgframe,RooFit.Components(",".join(full_bkg)),RooFit.Normalization(nd*(exp_event+adjust_norm), ROOT.RooAbsReal.NumEvent))
model_curve = bkgframe.getCurve()
fa_plus = np.array([model_curve.interpolate(x) for x in xvar])
bkgframe.remove(str(0),False)
fit_params.setRealValue(name, info['mean']-2.0*info['error'])
model.plotOn(bkgframe,RooFit.Components(",".join(full_bkg)),RooFit.Normalization(nd*(exp_event-adjust_norm), ROOT.RooAbsReal.NumEvent))
model_curve = bkgframe.getCurve()
fa_minus = np.array([model_curve.interpolate(x) for x in xvar])
bkgframe.remove(str(0),False)
if name == 'nsig':
fa.append(np.zeros(nbinx))
else:
fa.append((fa_plus - fa_minus) / (2.0*info['error']))
# reset the params matrix
fit_params.setRealValue(name, info['mean'])
fa = np.array(fa).T
tmp = np.array([np.asarray(np.matmul(FA, covMatrix)).flatten() for FA in fa])
bkg_unc = np.sqrt(np.array([np.dot(t, FA) for t, FA in zip(tmp, fa)]))
nbkg_total_err = np.sqrt(np.trapz(bkg_unc*bkg_unc, x=xvar)) / ((xmax-xmin)/nbin_data)
#print(np.trapz(model_cen, x=xvar) / ((xmax-xmin)/nbin_data))
fig, ax = plt.subplots()
ax.plot(xvar, model_cen, 'b-', label=r'$N_{{\rm bkg}}={0:.1f}\pm{1:.1f}$'.format(nbkg_total, nbkg_total_err))
ax.fill_between(xvar, model_cen-bkg_unc, model_cen+bkg_unc, facecolor='red', alpha=0.5, linewidth=0.0, label=r'$1\sigma$')
ax.set_xlabel(r'$m(K^{+}e^{+}e^{-}) [{\rm GeV}]$')
ax.set_ylabel(r'a.u.')
ax.set_ylim(bottom=0)
ax.legend(loc='upper right')
if plotSigmaBand:
fig.savefig(outputfile.replace('.pdf','')+'_totalbkg_1sigma.pdf', bbox_inches='tight')
#nbkg_total_err = 1.0
#nbkg_total_err = get_nbkg_err(xframe, model, partialfit, blinded, results, nbkg_total, plotSigmaBand, outputfile, sig_low, sig_up, xmin, xmax, nbin_data)
SNR = nsig_interested/np.sqrt(nsig_interested + nbkg_total)
print("Total number of background: {}, uncertainty: {}".format(nbkg_total, nbkg_total_err))
#print("S/sqrt(S+B): {}".format(nsig.getVal()/np.sqrt(nsig.getVal() + nbkg_total)))
print("S/sqrt(S+B): {}".format(SNR))
print("*"*80)
# Plot results of fit on a different frame
c2 = ROOT.TCanvas('fig_binnedFit', 'fit', 800, 600)
c2.SetGrid()
c2.cd()
ROOT.gPad.SetLeftMargin(0.10)
ROOT.gPad.SetRightMargin(0.05)
#xframe = wspace.var('x').frame(RooFit.Title("PF electron"))
xframe = theBMass.frame()
if isMC:
data.plotOn(xframe, RooFit.Binning(nbin_data), RooFit.Name("data"))
model.plotOn(xframe,RooFit.Name("global"),RooFit.Range("Full"),RooFit.LineColor(2),RooFit.MoveToBack()) # this will show fit overlay on canvas
if fitmode == 'b':
model.paramOn(xframe,RooFit.Layout(0.60,0.92,0.73))
else:
model.paramOn(xframe,RooFit.Layout(0.15,0.45,0.73))
xframe.getAttText().SetTextSize(0.03)
else:
if blinded:
#nd = data.reduce('(({0}_fit_mass > {1}) & ({0}_fit_mass < {2})) | (({0}_fit_mass > {3}) & ({0}_fit_mass < {4}))'.format(prefix, xmin, sig_low, sig_up, xmax)).sumEntries() / data.reduce('({0}_fit_mass > {1}) & ({0}_fit_mass < {2})'.format(prefix, xmin, xmax)).sumEntries()
nd = 1.0
data.plotOn(xframe, RooFit.Binning(nbin_data), RooFit.CutRange("SB1,SB2"), RooFit.Name("data"))
else:
nd = 1.0
data.plotOn(xframe, RooFit.Binning(nbin_data), RooFit.Name("data"))
model.plotOn(xframe,RooFit.Name("global"),RooFit.Range("Full"),RooFit.Normalization(nd, ROOT.RooAbsReal.Relative),RooFit.LineColor(2),RooFit.MoveToBack()) # this will show fit overlay on canvas
model.plotOn(xframe,RooFit.Name("bkg"),RooFit.Components("bkg"),RooFit.Range("Full"),RooFit.Normalization(nd, ROOT.RooAbsReal.Relative),RooFit.DrawOption("F"),RooFit.VLines(),RooFit.FillColor(42),RooFit.LineColor(42),RooFit.LineWidth(1),RooFit.MoveToBack())
plotted_partial = []
for name, info in partialfit.items():
model.plotOn(xframe,RooFit.Name(name),RooFit.Components("bkg,"+",".join(plotted_partial)+",{}".format(name)),RooFit.Range("Full"),RooFit.Normalization(nd, ROOT.RooAbsReal.Relative),RooFit.DrawOption("F"),RooFit.VLines(),RooFit.FillColor(info['color']),RooFit.LineColor(info['color']),RooFit.LineWidth(1),RooFit.MoveToBack())
plotted_partial.append(name)
model.plotOn(xframe,RooFit.Name("sig"),RooFit.Components("sig"),RooFit.Range("Full"),RooFit.Normalization(nd, ROOT.RooAbsReal.Relative),RooFit.DrawOption("L"),RooFit.LineStyle(2),RooFit.LineColor(1))
model.paramOn(xframe,RooFit.Layout(0.15,0.45,0.83))
xframe.getAttText().SetTextSize(0.03)
xframe.GetYaxis().SetTitleOffset(0.9)
xframe.GetYaxis().SetTitleFont(42)
xframe.GetYaxis().SetTitleSize(0.05)
xframe.GetYaxis().SetLabelSize(0.04)
xframe.GetYaxis().SetLabelFont(42)
xframe.GetXaxis().SetTitleOffset(0.9)
xframe.GetXaxis().SetTitleFont(42)
xframe.GetXaxis().SetTitleSize(0.05)
xframe.GetXaxis().SetLabelSize(0.04)
xframe.GetXaxis().SetLabelFont(42)
xframe.GetYaxis().SetTitle("Events / {0:.0f} MeV".format((xmax - xmin)/nbin_data*1000.))
xtitle = "m(K^{+}e^{+}e^{-}) [GeV]" if prefix == 'BToKEE' else "m(K^{+}K^{-}e^{+}e^{-}) [GeV]"
if fitmode == 'jpsi':
xtitle = "m(e^{+}e^{-}) [GeV]"
elif fitmode == 'phi':
xtitle = "m(K^{+}K^{-}) [GeV]"
xframe.GetXaxis().SetTitle(xtitle)
#xframe.GetXaxis().SetTitle("2nd principal component [GeV]" if fitJpsi else "m(K^{+}e^{+}e^{-}) [GeV]")
xframe.SetStats(0)
xframe.SetMinimum(0)
xframe.Draw()
CMS_lumi(isMC)
if isMC:
if fitmode == 'b':
legend = ROOT.TLegend(0.65,0.75,0.92,0.85);
pt = ROOT.TPaveText(0.72,0.38,0.92,0.50,"brNDC")
else:
legend = ROOT.TLegend(0.15,0.75,0.42,0.85);
pt = ROOT.TPaveText(0.15,0.38,0.45,0.50,"brNDC")
legend.AddEntry(xframe.findObject("global"),"Total Fit","l");
else:
legend = ROOT.TLegend(0.56,0.65,0.92,0.85) if prefix == 'BToKEE' else ROOT.TLegend(0.46,0.70,0.92,0.85)
legend.AddEntry(xframe.findObject("bkg"),"Combinatorial","f");
pt = ROOT.TPaveText(0.7,0.35,0.92,0.63,"brNDC")
#pt = ROOT.TPaveText(0.72,0.30,0.92,0.63,"brNDC")
for name, info in partialfit.items():
legend.AddEntry(xframe.findObject(name),info['label'],"f");
legend.AddEntry(xframe.findObject("sig"),sigName,"l");
legend.SetTextFont(42);
legend.SetTextSize(0.04);
legend.AddEntry(xframe.findObject("data"),"Data","lpe");
legend.Draw();
if drawSNR:
pt.SetFillColor(0)
pt.SetBorderSize(1)
pt.SetTextFont(42);
pt.SetTextSize(0.04);
pt.SetTextAlign(12)
pt.AddText("MVA cut: {0:.2f}".format(mvaCut))
pt.AddText("S_{{total}}: {0:.0f}#pm{1:.0f}".format(nsig.getVal(),nsig.getError()))
pt.AddText("S: {0:.0f}#pm{1:.0f}".format(nsig_interested,nsig_interested_err))
if not isMC:
pt.AddText("B: {0:.0f}#pm{1:.0f}".format(nbkg_total, nbkg_total_err))
#pt.AddText("S/#sqrt{{S+B}}: {0:.1f}".format(nsig.getVal()/np.sqrt(nsig.getVal() + nbkg_total)))
pt.AddText("S/#sqrt{{S+B}}: {0:.1f}".format(SNR))
#pt.AddText("Punzi: {0:.1f}".format(Punzi(nbkgWindow, 2.0, 5.0)))
if not fitConverged:
pt.AddText("Fit is not converged")
pt.Draw()
c2.cd()
c2.Update()
c2.SaveAs(outputfile.replace('.pdf','')+'.pdf')
print("="*80)
if not isMC:
#return nsig.getVal(), nsig.getError(), nbkg_total, nbkg_total_err
#return nsig.getVal(), nsig.getError(), nsig_interested, nsig_interested_err, nbkg_total, nbkg_total_err
output = {}
output['Stot'] = nsig.getVal()
output['StotErr'] = nsig.getError()
output['S'] = nsig_interested
output['SErr'] = nsig_interested_err
output['B'] = nbkg_total
output['BErr'] = nbkg_total_err
output['exp_alpha'] = fit_params.find('exp_alpha').getVal()
return output
else:
return 0.0, 0.0, 0.0, 0.0, 0.0, 0.0
type=str,
default='1e-10',
help='Choose tolerance.')
args = parser.parse_args()
iteration = args.iteration
tolerance = args.tolerance
# open the input file
infn = 'stored_matrices/{}/zero_suppressed_matrix_iter{}.root'.format(
tolerance, iteration)
tinf = TFile(infn)
# tinf.ls()
# retrieve the zero suppressed covariance matrix and cast it into a numpy matrix
tcov_zero_sup = tinf.Get('cov_zero_sup')
cov_mat = matrix(tcov_zero_sup)
# results container
cosines = []
# loop through all columns
for i in range(cov_mat.shape[1] - 1):
# get the adjacent column vectors
v1 = cov_mat[:, i]
v2 = cov_mat[:, i + 1]
cosines.append(cosine_between(v1, v2))
# make plot
bins = np.linspace(-1.0, 1.0, 51)
plt.figure(figsize=(5, 5))
plt.hist(cosines, bins=bins, histtype='step')
plt.xlabel(r'$\cos(\mathbf{v}_i,\mathbf{v}_{i+1})$')