def test_comp_2dSymfit():
logger.info('Test multi-component 2d Sym fit')
model = Models.Fit2DSym(name='fit_comp',
signal_x=signal_x1,
signal_y=signal_y1,
bkg_1x=bkg_x,
components=[ss_cmp, sb_cmp, bb_cmp])
with rooSilent():
## components
model.SS.fix(5000)
model.SB.fix(1000)
model.BB.fix(1000)
model.C[0].fix(5000)
model.C[1].fix(1000)
model.C[2].fix(1000)
r = model.fitTo(dataset, ncpu=8)
model.SS.release()
model.SB.release()
model.BB.release()
model.C[0].release()
model.C[1].release()
model.C[2].release()
r = model.fitTo(dataset, ncpu=8)
logger.info('Model %s Fit result \n#%s ' % (model.name, r))
def test_psxps_BBsym () :
logger.info ('Simmetric fit model with non-factorizeable background component: ( Gauss + P1 ) (x) ( Gauss + P1 ) + (PS*P1)**2')
PS = Ostap.Math.PhaseSpaceNL( 1.0 , 5.0 , 2 , 5 )
model = Models.Fit2DSym (
suffix = '_12' ,
signal_x = signal1 ,
signal_y = signal2s ,
bkg_1x = -1 ,
bkg_2D = Models.PSPol2Dsym_pdf ( 'P2D12' , m_x , m_y , ps = PS , n = 1 )
)
## fit with fixed mass and sigma
with rooSilent() :
result, frame = model. fitTo ( dataset )
model.signal_x.sigma.release ()
model.signal_y.sigma.release ()
model.signal_x.mean .release ()
model.signal_y.mean .release ()
result, frame = model. fitTo ( dataset )
if 0 != result.status() or 3 != result.covQual() :
logger.warning('Fit is not perfect MIGRAD=%d QUAL=%d '
% ( result.status() , result.covQual() ) )
print(result)
else :
logger.info ('S1xS2 : %20s' % result ( model.SS ) [0] )
logger.info ('S1xB2 : %20s' % ( result ( model.SB ) [0] /2 ) )
logger.info ('B1xS2 : %20s' % ( result ( model.BS ) [0] /2 ) )
logger.info ('B1xB2 : %20s' % result ( model.BB ) [0] )
models.add ( model )
def test_p1xp1_BBsym():
logger.info(
'Symmetric non-factorized fit model: ( Gauss + P1 ) (x) ( Gauss + P1 ) + BBsym'
)
sb = ROOT.RooRealVar('sb', 'SB', 0, 10000)
model = Models.Fit2DSym(
suffix='_6',
signal_x=signal1,
signal_y=signal2s,
bkg_1x=-1,
bkg_2D=Models.PolyPos2Dsym_pdf('P2D5', m_x, m_y, n=2),
)
## fit with fixed mass and sigma
with rooSilent():
result, frame = model.fitTo(dataset)
model.signal_x.sigma.release()
model.signal_y.sigma.release()
model.signal_x.mean.release()
model.signal_y.mean.release()
result, frame = model.fitTo(dataset)
model.draw1(dataset)
model.draw2(dataset)
if 0 != result.status() or 3 != result.covQual():
logger.warning('Fit is not perfect MIGRAD=%d QUAL=%d ' %
(result.status(), result.covQual()))
print(result)
else:
logger.info('S1xS2 : %20s' % result(model.SS)[0])
logger.info('S1xB2 : %20s' % (result(model.SB)[0] / 2))
logger.info('B1xS2 : %20s' % (result(model.BS)[0] / 2))
logger.info('B1xB2 : %20s' % result(model.BB)[0])
models.add(model)
def test_model_16():
logger.info(
'Symmetric fit model with non-factorazeable symmetric (spline) background component: ( Gauss + expo*P1 ) (x) ( Gauss + expo*P1 ) + Spline2Dsym'
)
SPLINES = Ostap.Math.PositiveSpline2DSym(spline1)
model = Models.Fit2DSym(suffix='_16',
signal_x=signal1,
signal_y=signal2s,
bkg_1x=1,
bkg_2D=Models.Spline2Dsym_pdf('P2D16',
m_x,
m_y,
spline=SPLINES))
## fit with fixed mass and sigma
with rooSilent():
result, frame = model.fitTo(dataset)
model.signal_x.sigma.release()
model.signal_y.sigma.release()
model.signal_x.mean.release()
model.signal_y.mean.release()
result, frame = model.fitTo(dataset)
if 0 != result.status() or 3 != result.covQual():
logger.warning('Fit is not perfect MIGRAD=%d QUAL=%d ' %
(result.status(), result.covQual()))
print(result)
else:
logger.info('S1xS2 : %20s' % result(model.SS)[0])
logger.info('S1xB2 : %20s' % (result(model.SB)[0] / 2))
logger.info('B1xS2 : %20s' % (result(model.BS)[0] / 2))
logger.info('B1xB2 : %20s' % result(model.BB)[0])
models.add(model)
def test_pbxpb_BBsym():
logger.info(
'Symmetric fit model with non-factorizeable background component: ( Gauss + P1 ) (x) ( Gauss + P1 ) + Sym(expo*P1)**2'
)
model = Models.Fit2DSym(suffix='_9',
signal_1=signal1,
signal_2=signal2s,
bkg1=-1,
bkg2D=Models.ExpoPol2Dsym_pdf('P2D9',
m_x,
m_y,
n=1))
## fit with fixed mass and sigma
with rooSilent():
result, frame = model.fitTo(dataset)
model.signal1.sigma.release()
model.signal2.sigma.release()
model.signal1.mean.release()
model.signal2.mean.release()
result, frame = model.fitTo(dataset)
if 0 != result.status() or 3 != result.covQual():
logger.warning('Fit is not perfect MIGRAD=%d QUAL=%d ' %
(result.status(), result.covQual()))
print result
else:
logger.info('S1xS2 : %20s' % result(model.SS)[0])
logger.info('S1xB2 : %20s' % (result(model.SB)[0] / 2))
logger.info('B1xS2 : %20s' % (result(model.BS)[0] / 2))
logger.info('B1xB2 : %20s' % result(model.BB)[0])
models.add(model)
def test_psxps_BBsym():
logger.info(
'Simmetric fit model with non-factorizeable background component: ( Gauss + expo*P1 ) (x) ( Gauss + expo*P1 ) + (PS*P1)**2'
)
PS = Ostap.Math.PhaseSpaceNL(1.0, 5.0, 2, 5)
model = Models.Fit2DSym(suffix='_12',
signal_1=Models.Gauss_pdf('Gx',
m_x.getMin(),
m_x.getMax(),
mass=m_x),
signal_2=Models.Gauss_pdf('Gy',
m_y.getMin(),
m_y.getMax(),
mass=m_y),
bkg1=1,
bkg2D=Models.PSPol2Dsym_pdf('P2D12',
m_x,
m_y,
ps=PS,
n=1))
model.signal1.sigma.fix(m.error())
model.signal2.sigma.fix(m.error())
model.signal1.mean.fix(m.value())
model.signal2.mean.fix(m.value())
model.signal1.mean.fix(m.value())
model.signal2.mean.fix(m.value())
model.bkg1.tau.fix(0)
model.bkg2.tau.fix(0)
## fit with fixed mass and sigma
with rooSilent():
result, frame = model.fitTo(dataset)
model.signal1.sigma.release()
model.signal2.sigma.release()
model.signal1.mean.release()
model.signal2.mean.release()
result, frame = model.fitTo(dataset)
if 0 != result.status() or 3 != result.covQual():
logger.warning('Fit is not perfect MIGRAD=%d QUAL=%d ' %
(result.status(), result.covQual()))
print result
else:
logger.info('S1xS2 : %20s' % result(model.ss)[0])
logger.info('S1xB2 : %20s' % (result(model.sb)[0] / 2))
logger.info('B1xS2 : %20s' % (result(model.bs)[0] / 2))
logger.info('B1xB2 : %20s' % result(model.bb)[0])
models.add(model)
def test_model_16():
logger.info(
'Symmetric fit model with non-factorazeable symmetric (spline) backgrond component: ( Gauss + expo*P1 ) (x) ( Gauss + expo*P1 ) + Spline2Dsym'
)
SPLINES = Ostap.Math.Spline2DSym(spline1)
model = Models.Fit2DSym(suffix='_16',
signal_1=Models.Gauss_pdf('Gx',
m_x.getMin(),
m_x.getMax(),
mass=m_x),
signal_2=Models.Gauss_pdf('Gy',
m_y.getMin(),
m_y.getMax(),
mass=m_y),
bkg1=1,
bkg2D=Models.Spline2Dsym_pdf('P2D16',
m_x,
m_y,
spline=SPLINES))
model.signal1.sigma.fix(m.error())
model.signal2.sigma.fix(m.error())
model.signal1.mean.fix(m.value())
model.signal2.mean.fix(m.value())
model.signal1.mean.fix(m.value())
model.signal2.mean.fix(m.value())
model.bkg1.tau.fix(0)
## fit with fixed mass and sigma
with rooSilent():
result, frame = model.fitTo(dataset)
model.signal1.sigma.release()
model.signal2.sigma.release()
model.signal1.mean.release()
model.signal2.mean.release()
result, frame = model.fitTo(dataset)
if 0 != result.status() or 3 != result.covQual():
logger.warning('Fit is not perfect MIGRAD=%d QUAL=%d ' %
(result.status(), result.covQual()))
print result
else:
logger.info('S1xS2 : %20s' % result(model.ss)[0])
logger.info('S1xB2 : %20s' % (result(model.sb)[0] / 2))
logger.info('B1xS2 : %20s' % (result(model.bs)[0] / 2))
logger.info('B1xB2 : %20s' % result(model.bb)[0])
models.add(model)
def test_p1xp1_BBsym():
logger.info(
'Symmetric non-factorized fit model: ( Gauss + P1 ) (x) ( Gauss + P1 ) + BBsym'
)
sb = ROOT.RooRealVar('sb', 'SB', 0, 10000)
model = Models.Fit2DSym(
suffix='_6',
signal_1=Models.Gauss_pdf('Gx', m_x.getMin(), m_x.getMax(), mass=m_x),
signal_2=Models.Gauss_pdf('Gy', m_y.getMin(), m_y.getMax(), mass=m_y),
bkg1=1,
bkg2D=Models.PolyPos2Dsym_pdf('P2D5', m_x, m_y, n=2),
)
model.signal1.sigma.fix(m.error())
model.signal2.sigma.fix(m.error())
model.signal1.mean.fix(m.value())
model.signal2.mean.fix(m.value())
model.signal1.mean.fix(m.value())
model.signal2.mean.fix(m.value())
model.bkg1.tau.fix(0)
model.bkg2.tau.fix(0)
## fit with fixed mass and sigma
with rooSilent():
result, frame = model.fitTo(dataset)
model.signal1.sigma.release()
model.signal2.sigma.release()
model.signal1.mean.release()
model.signal2.mean.release()
result, frame = model.fitTo(dataset)
if 0 != result.status() or 3 != result.covQual():
logger.warning('Fit is not perfect MIGRAD=%d QUAL=%d ' %
(result.status(), result.covQual()))
print result
else:
logger.info('S1xS2 : %20s' % result(model.ss)[0])
logger.info('S1xB2 : %20s' % (result(model.sb)[0] / 2))
logger.info('B1xS2 : %20s' % (result(model.bs)[0] / 2))
logger.info('B1xB2 : %20s' % result(model.bb)[0])
models.add(model)
def test_simfit3():
logger = getLogger('test_simfit3')
## low statistic, high-background "signal channel"
signal2 = Models.Gauss_pdf('G2',
xvar=mass2,
mean=(15, 20),
sigma=(0.1, 5.0))
signal3 = Models.Gauss_pdf('G3',
xvar=mass3,
mean=signal2.mean,
sigma=signal2.sigma)
model_S = Models.Fit2DSym(suffix='M2',
yvar=mass3,
signal_x=signal2,
signal_y=signal3,
bkg_1x=None,
bkg_2x=None)
model_S.BB = NB2
model_S.SS = NS2
model_S.SB = NC2
model_S.SB.setMin(-500)
## fit 2
rS, fx = model_S.fitTo(dataset2, draw=None, nbins=50, silent=True)
rS, fx = model_S.fitTo(dataset2, draw=None, nbins=50, silent=True)
title = 'Results of fit to signal sample only'
logger.info('%s\n%s' % (title, rS.table(title=title, prefix='# ')))
with use_canvas('test_simfit3'):
with wait(1):
rS, fx = model_S.fitTo(dataset2, draw='X', nbins=50, silent=True)
with wait(1):
rS, fy = model_S.fitTo(dataset2, draw='Y', nbins=50, silent=True)
# =========================================================================
## high statistic, low-background "control/normalization channel"
mean_N = signal2.vars_subtract(signal2.mean, 10.0)
sigma_N = signal2.vars_multiply(signal2.sigma, 0.5)
signal_N = Models.Gauss_pdf('G1', xvar=mass1, mean=mean_N, sigma=sigma_N)
model_N = Models.Fit1D(suffix='M1', signal=signal_N, background=-1)
model_N.S = NS1
model_N.B = NB1
## fit 1
rN, fN = model_N.fitTo(dataset1, draw=False, nbins=50, silent=True)
with use_canvas('test_simfit3'), wait(1):
rN, fN = model_N.fitTo(dataset1, draw=True, nbins=50, silent=True)
title = 'Results of fit to normalization sample only'
logger.info('%s\n%s' % (title, rN.table(title=title, prefix='# ')))
# =========================================================================
## combine data
sample = ROOT.RooCategory('sample', 'sample', 'S', 'N')
## combine datasets
from ostap.fitting.simfit import combined_data
vars = ROOT.RooArgSet(mass1, mass2, mass3)
dataset = combined_data(sample, vars, {'N': dataset1, 'S': dataset2})
## combine PDFs
model_sim = Models.SimFit(sample, {'S': model_S, 'N': model_N}, name='X')
# =========================================================================
rC, fC = model_sim.fitTo(dataset, silent=True)
rC, fC = model_sim.fitTo(dataset, silent=True)
title = 'Results of simultaneous fit'
logger.info('%s\n%s' % (title, rC.table(title=title, prefix='# ')))
with use_canvas('test_simfit3'), wait(1):
with wait(1):
fN = model_sim.draw('N', dataset, nbins=50)
with wait(1):
fSx = model_sim.draw('S/x', dataset, nbins=50)
with wait(1):
fSy = model_sim.draw('S/y', dataset, nbins=50)
logger.info(' Value | Simple fit | Combined fit ')
logger.info(' #N | %25s | %-25s ' % (rS.SSM2 * 1, rC.SSM2 * 1))
logger.info(' mean | %25s | %-25s ' % (rS.mean_G2 * 1, rC.mean_G2 * 1))
logger.info(' sigma | %25s | %-25s ' % (rS.sigma_G2 * 1, rC.sigma_G2 * 1))
# =============================================================================
## low statistic, high-background "signal channel"
signal2 = Models.Gauss_pdf ( 'G2' ,
xvar = mass2 ,
mean = ( 15 , 20 ) ,
sigma = ( 0.1 , 5.0 ) )
signal3 = Models.Gauss_pdf ( 'G3' ,
xvar = mass3 ,
mean = signal2.mean ,
sigma = signal2.sigma )
model_S = Models.Fit2DSym ( suffix = 'M2' ,
yvar = mass3 ,
signal_x = signal2 ,
signal_y = signal3 ,
bkg_1x = None ,
bkg_2x = None )
model_S.BB = NB2
model_S.SS = NS2
model_S.SB = NC2
model_S.SB.setMin ( -500 )
## fit 2
rS , fx = model_S.fitTo ( dataset2 , draw = None , nbins = 50 , silent = True )
rS , fx = model_S.fitTo ( dataset2 , draw = None , nbins = 50 , silent = True )
rS , fx = model_S.fitTo ( dataset2 , draw = 'X' , nbins = 50 , silent = True )
rS , fy = model_S.fitTo ( dataset2 , draw = 'Y' , nbins = 50 , silent = True )
logger.info ( 'Fit results for signal sample only: %s' % rS )
