def minimize(self):
w = self.model.w
nll = w.pdf('model').createNLL(w.data('data'), *self.fitArgs[:-1])
minu = r.RooMinuit(nll)
minu.setPrintLevel(-1)
minu.setNoWarn()
for j in range(10):
minu.setStrategy(2)
for i in range(10):
self.fitstatus = minu.migrad()
print>>self.log, i + 10*j,
self.log.flush()
if not self.fitstatus: break
if not self.fitstatus: break
minu.setStrategy(1)
minu.migrad()
print>>self.log
self.NLL = nll.getVal()
pll = nll.createProfile(w.argSet(','.join(self.profileVars)))
print>>self.log, roo.str(nll)
print>>self.log, roo.str(pll)
if hasattr(pll,'minimizer'):
pll.minimizer().setStrategy(2)
else: pll.minuit().setStrategy(2)
return pll
def minimize(self):
w = self.model.w
nll = w.pdf('model').createNLL(w.data('data'), *self.fitArgs[:-1])
minu = r.RooMinuit(nll)
minu.setPrintLevel(-1)
minu.setNoWarn()
for j in range(10):
minu.setStrategy(2)
for i in range(10):
self.fitstatus = minu.migrad()
print >> self.log, i + 10 * j,
self.log.flush()
if not self.fitstatus: break
if not self.fitstatus: break
minu.setStrategy(1)
minu.migrad()
print >> self.log
self.NLL = nll.getVal()
pll = nll.createProfile(w.argSet(','.join(self.profileVars)))
print >> self.log, roo.str(nll)
print >> self.log, roo.str(pll)
if hasattr(pll, 'minimizer'):
pll.minimizer().setStrategy(2)
else:
pll.minuit().setStrategy(2)
return pll
def minimize(self):
w = self.model.w
nll = w.pdf(self.nobg+'model'+self.only).createNLL(w.data('data'+self.only), *self.fitArgs[:-1])
minu = r.RooMinuit(nll)
minu.setPrintLevel(-1)
minu.setNoWarn()
for j in range(10):
minu.setStrategy(2)
for i in range(10):
self.fitstatus = minu.migrad()
print>>self.log, i + 10*j,
self.log.flush()
if not self.fitstatus: break
if not self.fitstatus: break
minu.setStrategy(1)
minu.migrad()
print>>self.log
print>>self.log, roo.str(w.arg('alpha'))
self.NLL = nll.getVal()
alpha = w.arg('alpha').getVal(), w.arg('alpha').getError()
pll = nll.createProfile(w.argSet('alpha'))
print>>self.log, roo.str(nll)
print>>self.log, roo.str(pll)
if hasattr(pll,'minimizer'):
pll.minimizer().setStrategy(2)
else: pll.minuit().setStrategy(2)
return alpha,pll
def doFit(self):
w = self.model.w
contourPoints = None
while not contourPoints:
pll = self.minimize()
p0, contourPoints = self.contourPoints(pll)
ell = enclosing_ellipse([p[:2] for p in contourPoints], p0[:2])
self.profVal = p0[:2]
self.profErr = ell.sigmas2
self.profPLL = p0[2]
self.scales = np.array(
[w.arg(a).getVal() for a in ['Ac_y_ttqq', 'Ac_y_ttqg']])
self.scalesPhi = [
w.arg('Ac_phi_%s' % n).getVal()
for n in ['ttqq', 'ttgg', 'ttag', 'ttqg', 'tt']
]
self.correction = w.arg('Ac_y_ttgg').getVal() * w.arg('f_gg').getVal()
self.fractionHats = [
w.arg('f_%s_hat' % a).getVal() for a in ['gg', 'qg', 'qq', 'ag']
]
fitXY = self.profVal * self.scales
sigmas2 = np.diag(self.scales).dot(self.profErr).dot(
np.diag(self.scales))
self.fitX, self.fitY = [float(i) for i in fitXY]
self.fitXX = float(sigmas2[0, 0])
self.fitXY = float(sigmas2[0, 1])
self.fitYY = float(sigmas2[1, 1])
self.sigmaX = math.sqrt(self.fitXX / (2 * oneSigmaNLL))
self.sigmaY = math.sqrt(self.fitYY / (2 * oneSigmaNLL))
self.contourPointsX, self.contourPointsY, = zip(
*[[float(i) for i in self.scales * p] for p in contourPoints])
if not self.quiet:
print >> self.log, self.profVal, self.profPLL
print >> self.log, self.profErr
for item in [
'd_qq', 'd_xs_tt', 'd_xs_wj', 'factor_elqcd',
'factor_muqcd', 'f_gg', 'f_qq', 'f_qg', 'f_ag', 'R_ag',
'slosh', 'alphaL', 'alphaT'
]:
print >> self.log, '\t', roo.str(w.arg(item))
self.pll = pll
w.arg('falphaL').setVal(p0[0])
w.arg('falphaT').setVal(p0[1])
pll.getVal()
return
def doFit(self):
w = self.model.w
(alpha,alphaE), pll = self.minimize()
def pllEval(p):
w.arg('alpha').setVal(p)
return pll.getVal()
step = 0.2 * alphaE
points = [alpha+i*step for i in range(-10,11)]
pllPoints = [pllEval(p) for p in points]
iL = min((abs(1-v),i) for i,v in list(enumerate(pllPoints))[:10])[1]
iR = min((abs(1-v),i) for i,v in list(enumerate(pllPoints))[10:])[1]
iBounds = (iL, iR)
parb = parabola([(points[i],pllPoints[i]) for i in (10,)+iBounds])
self.profVal = float(parb.xmin)
self.profErr = parb.dx(0.5)
self.scale = w.arg('Ac_tt').getVal()
self.fit = self.profVal * self.scale
self.sigma = self.profErr * self.scale
self.profPLL = pllEval(self.profVal)
for item in ['pll','points','pllPoints']: setattr(self,item,eval(item))
self.parbABC = list(parb.ABC)
if not self.quiet:
print>>self.log
print>>self.log, 'Ac: %f +/- %f'% (self.fit, self.sigma)
print>>self.log
print>>self.log, "fit alpha: %f +/- %f"%(alpha,alphaE)
print>>self.log, "parb alpha: %f +/- %f"%(self.profVal, self.profErr)
print>>self.log, "min PLL:", self.profPLL
print>>self.log, parb
for item in ['d_xs_tt','d_xs_wj','factor_elqcd','factor_muqcd']:
print>>self.log, '\t', roo.str(w.arg(item))
print>>self.log
print>>self.log, 'iBounds', iBounds
for i,(p,v) in enumerate(zip(points,pllPoints)):
print>>self.log, i, p, v
w.arg('alpha').setError(self.profErr)
return
def doFit(self):
w = self.model.w
contourPoints=None
while not contourPoints:
pll = self.minimize()
p0, contourPoints = self.contourPoints(pll)
ell = enclosing_ellipse([p[:2] for p in contourPoints],p0[:2])
self.profVal = p0[:2]
self.profErr = ell.sigmas2
self.profPLL = p0[2]
self.scales = np.array([w.arg(a).getVal() for a in ['Ac_y_ttqq', 'Ac_y_ttqg']])
self.scalesPhi= [w.arg('Ac_phi_%s'%n).getVal() for n in ['ttqq','ttgg','ttag','ttqg','tt']]
self.correction = w.arg('Ac_y_ttgg').getVal() * w.arg('f_gg').getVal()
self.fractionHats = [w.arg('f_%s_hat' % a).getVal() for a in ['gg','qg','qq','ag']]
fitXY = self.profVal*self.scales
sigmas2 = np.diag(self.scales).dot(self.profErr).dot(np.diag(self.scales))
self.fitX,self.fitY = [float(i) for i in fitXY]
self.fitXX = float(sigmas2[0,0])
self.fitXY = float(sigmas2[0,1])
self.fitYY = float(sigmas2[1,1])
self.sigmaX = math.sqrt( self.fitXX / (2*oneSigmaNLL))
self.sigmaY = math.sqrt( self.fitYY / (2*oneSigmaNLL))
self.contourPointsX,self.contourPointsY, = zip(*[[float(i) for i in self.scales*p] for p in contourPoints])
if not self.quiet:
print>>self.log, self.profVal, self.profPLL
print>>self.log, self.profErr
for item in ['d_qq','d_xs_tt','d_xs_wj',
'factor_elqcd','factor_muqcd',
'f_gg','f_qq','f_qg','f_ag',
'R_ag','slosh','alphaL','alphaT']:
print>>self.log, '\t', roo.str(w.arg(item))
self.pll = pll
w.arg('falphaL').setVal(p0[0])
w.arg('falphaT').setVal(p0[1])
pll.getVal()
return
