def __addParameter(self, extrapars, extraparerrors, parnames, ndata,
ncorrsyst, errorkey, parindxmaps, ierr):
extrapars.append(0.0)
extraparerrors.append(1.0)
parnames.append(stripLeadingDigits(errorkey))
indxmap = {}
for ival in range(ndata):
indxmap[ival] = ncorrsyst
parindxmaps[ierr] = indxmap
return
def __addParameter( self, extrapars, extraparerrors, parnames,
ndata, ncorrsyst, errorkey, parindxmaps, ierr ):
extrapars.append( 0.0 )
extraparerrors.append( 1.0 )
parnames.append( stripLeadingDigits( errorkey ) )
indxmap= {}
for ival in range( ndata ):
indxmap[ival]= ncorrsyst
parindxmaps[ierr]= indxmap
return
def __createExtraPars(self):
dataparser = self._getDataparser()
herrors = dataparser.getErrors()
correlations = dataparser.getCorrelations()
hcovopt = dataparser.getCovoption()
extrapars = []
extraparerrors = []
extraparnames = []
errorkeys = sorted(herrors.keys())
ncorrsyst = 0
parindxmaps = {}
for errorkey in errorkeys:
ierr = errorkeys.index(errorkey)
errlist = herrors[errorkey]
ndata = len(errlist)
covopt = hcovopt[errorkey]
# Fully or globally partially correlated:
# add measured pseudo-parameter:
if ("f" in covopt or "gp" in covopt):
self.__addParameter(extrapars, extraparerrors, extraparnames,
ndata, ncorrsyst, errorkey, parindxmaps,
ierr)
ncorrsyst += 1
# Get correlations from matrix, add measured pseudo-
# parameter for each independent group of variables as
# detected by equal matrix row patterns.
elif "m" in covopt:
covoptlist = correlations[errorkey]
nsq = len(covoptlist)
dim = int(sqrt(nsq))
covoptmatrix = [
covoptlist[dim * i:dim * (i + 1)] for i in range(dim)
]
if (("f" in covoptlist or "gp" in covoptlist)
and not "p" in covoptlist):
rowpatterns = []
indxmap = {}
for row in covoptmatrix:
if row not in rowpatterns:
rowpatterns.append(row)
extrapars.append(0.0)
extraparerrors.append(1.0)
extraparname = stripLeadingDigits(errorkey)
valuenumbers = ""
for icovopt in range(len(row)):
if "f" in row[icovopt] or "gp" in row[icovopt]:
indxmap[icovopt] = ncorrsyst
valuenumbers += str(icovopt)
if len(valuenumbers) > 0:
extraparname += "_" + valuenumbers
extraparnames.append(extraparname)
ncorrsyst += 1
parindxmaps[ierr] = indxmap
return extrapars, extraparerrors, extraparnames, parindxmaps, errorkeys
def __createExtraPars( self ):
dataparser= self._getDataparser()
herrors= dataparser.getErrors()
correlations= dataparser.getCorrelations()
hcovopt= dataparser.getCovoption()
extrapars= []
extraparerrors= []
extraparnames= []
errorkeys= sorted( herrors.keys() )
ncorrsyst= 0
parindxmaps= {}
for errorkey in errorkeys:
ierr= errorkeys.index( errorkey )
errlist= herrors[errorkey]
ndata= len(errlist)
covopt= hcovopt[errorkey]
# Fully or globally partially correlated:
# add measured pseudo-parameter:
if( "f" in covopt or "gp" in covopt ):
self.__addParameter( extrapars, extraparerrors, extraparnames,
ndata, ncorrsyst, errorkey,
parindxmaps, ierr )
ncorrsyst+= 1
# Get correlations from matrix, add measured pseudo-
# parameter for each independent group of variables as
# detected by equal matrix row patterns.
elif "m" in covopt:
covoptlist= correlations[errorkey]
nsq= len(covoptlist)
dim= int(sqrt(nsq))
covoptmatrix= [ covoptlist[dim*i:dim*(i+1)]
for i in range(dim) ]
if( ( "f" in covoptlist or "gp" in covoptlist ) and
not "p" in covoptlist ):
rowpatterns= []
indxmap= {}
for row in covoptmatrix:
if row not in rowpatterns:
rowpatterns.append( row )
extrapars.append( 0.0 )
extraparerrors.append( 1.0 )
extraparname= stripLeadingDigits( errorkey )
valuenumbers= ""
for icovopt in range(len(row)):
if "f" in row[icovopt] or "gp" in row[icovopt]:
indxmap[icovopt]= ncorrsyst
valuenumbers+= str(icovopt)
if len(valuenumbers) > 0:
extraparname+= "_" + valuenumbers
extraparnames.append( extraparname )
ncorrsyst+= 1
parindxmaps[ierr]= indxmap
return extrapars, extraparerrors, extraparnames, parindxmaps, errorkeys
def printInputs( self, printcovopt=False ):
print "\n Best Linear Unbiased Estimator average"
self.dataparser.printInputs()
if printcovopt:
print "\n Covariance matrices:"
for key in sorted( self.hcov.keys() ):
print "{0:>10s}:".format( stripLeadingDigits( key ) )
self.__printMatrix( self.hcov[key] )
print "Total covariance:"
self.__printMatrix( self.cov )
corr= numpy.matrix( self.cov )
for i in range( corr.shape[0] ):
for j in range( corr.shape[1] ):
corr[i,j]= self.cov[i,j]/sqrt( self.cov[i,i]*self.cov[j,j] )
print "Total correlation:"
self.__printMatrix( corr, "6.3f" )
print "Inverse:"
self.__printMatrix( self.inv )
return
def printInputs(self, printcovopt=False):
print "\n Best Linear Unbiased Estimator average"
self.dataparser.printInputs()
if printcovopt:
print "\n Covariance matrices:"
for key in sorted(self.hcov.keys()):
print "{0:>10s}:".format(stripLeadingDigits(key))
self.__printMatrix(self.hcov[key])
print "Total covariance:"
self.__printMatrix(self.cov)
corr = numpy.matrix(self.cov)
for i in range(corr.shape[0]):
for j in range(corr.shape[1]):
corr[i, j] = self.cov[i, j] / sqrt(
self.cov[i, i] * self.cov[j, j])
print "Total correlation:"
self.__printMatrix(corr, "6.3f")
print "Inverse:"
self.__printMatrix(self.inv)
return
def printErrorsAndWeights(self, optinfo=False):
errors, weightsmatrix = self.errorAnalysis()
navg = weightsmatrix.shape[0]
nval = weightsmatrix.shape[1]
print "Error composition:"
if optinfo and navg == 1:
print " +/- errors dI/df/I offd. sums"
hinfos, hinfosums = self.informationAnalysis(weightsmatrix)
errorkeys = sorted(errors.keys())
errorkeys.remove("totalcov")
errorkeys.remove("systcov")
for errorkey in errorkeys:
print "{0:>10s}:".format(stripLeadingDigits(errorkey)),
for iavg in range(navg):
error = errors[errorkey]
print "{0:10.4f}".format(sqrt(error[iavg, iavg])),
if navg == 1 and optinfo and not ("syst" in errorkey
or "total" in errorkey):
print "{0:9.3f}".format(hinfosums[errorkey]),
print
names = self.__dataparser.getNames()
print "\n Variables:",
for name in names:
print "{0:>10s}".format(name),
print
groups = sorted(set(self.__dataparser.getGroups()))
for iavg in range(navg):
txt = "Weights"
if navg > 1:
txt += " " + str(groups[iavg])
print "{0:>10s}:".format(txt),
for ival in range(nval):
print "{0:10.4f}".format(float(weightsmatrix[iavg, ival])),
print
if optinfo:
print " DeltaI/I:",
totalerrors = self.__dataparser.getTotalErrors()
for iavg in range(navg):
if iavg > 0:
print " ",
deltaIsum = 0.0
for ival in range(nval):
deltaI = errors["total"][iavg, iavg] / totalerrors[ival]**2
deltaIsum += deltaI
print "{0:10.4f}".format(deltaI),
print "{0:10.4f}".format(1.0 - deltaIsum)
print " Pulls:",
pulls = self.calcPulls()
for ival in range(nval):
print "{0:10.4f}".format(pulls[ival, 0]),
print
if navg > 1:
print "\nCorrelations:"
totcov = errors["total"]
for iavg in range(navg):
for javg in range(navg):
corr = totcov[iavg, javg] / sqrt(
totcov[iavg, iavg] * totcov[javg, javg])
print "{0:6.3f}".format(corr),
print
elif optinfo:
print "\n dI/df/I offdiagonals per error source:"
keys = sorted(hinfos.keys())
#keys.remove( "01stat" )
keys = [key for key in keys if not "stat" in key]
for key in keys:
print "{0:>10s}:".format(stripLeadingDigits(key))
infom = hinfos[key]
print " ",
for name in names[1:]:
print "{0:>7s}".format(name),
print
for i in range(nval - 1):
for j in range(nval):
if j == 0 and i < nval - 1:
print "{0:>7s}".format(names[i]),
elif j > i:
print "{0:7.4f}".format(infom[i, j]),
else:
print " ",
print
print
return
def test_stripLeadingDigits(self):
word = "01abc1"
expectedword = "abc1"
strippedword = stripLeadingDigits(word)
self.assertEqual(strippedword, expectedword)
return
def test_stripLeadingDigits( self ):
word= "01abc1"
expectedword= "abc1"
strippedword= stripLeadingDigits( word )
self.assertEqual( strippedword, expectedword )
return
def printErrorsAndWeights( self, optinfo=False ):
errors, weightsmatrix= self.errorAnalysis()
navg= weightsmatrix.shape[0]
nval= weightsmatrix.shape[1]
print "Error composition:"
if optinfo and navg == 1:
print " +/- errors dI/df/I offd. sums"
hinfos, hinfosums= self.informationAnalysis( weightsmatrix )
errorkeys= sorted( errors.keys() )
errorkeys.remove( "totalcov" )
errorkeys.remove( "systcov" )
for errorkey in errorkeys:
print "{0:>10s}:".format( stripLeadingDigits( errorkey ) ),
for iavg in range( navg ):
error= errors[errorkey]
print "{0:10.4f}".format( sqrt(error[iavg,iavg]) ),
if navg == 1 and optinfo and not ( "syst" in errorkey or
"total" in errorkey ):
print "{0:9.3f}".format( hinfosums[errorkey] ),
print
names= self.__dataparser.getNames()
print "\n Variables:",
for name in names:
print "{0:>10s}".format( name ),
print
groups= sorted(set(self.__dataparser.getGroups()))
for iavg in range( navg ):
txt= "Weights"
if navg > 1:
txt+= " "+str(groups[iavg])
print "{0:>10s}:".format( txt ),
for ival in range( nval ):
print "{0:10.4f}".format( float(weightsmatrix[iavg,ival]) ),
print
if optinfo:
print " DeltaI/I:",
totalerrors= self.__dataparser.getTotalErrors()
for iavg in range( navg ):
if iavg > 0:
print " ",
deltaIsum= 0.0
for ival in range( nval ):
deltaI= errors["total"][iavg,iavg]/totalerrors[ival]**2
deltaIsum+= deltaI
print "{0:10.4f}".format( deltaI ),
print "{0:10.4f}".format( 1.0-deltaIsum )
print " Pulls:",
pulls= self.calcPulls()
for ival in range( nval ):
print "{0:10.4f}".format( pulls[ival,0] ),
print
if navg > 1:
print "\nCorrelations:"
totcov= errors["total"]
for iavg in range( navg ):
for javg in range( navg ):
corr= totcov[iavg,javg]/sqrt(totcov[iavg,iavg]*totcov[javg,javg])
print "{0:6.3f}".format( corr ),
print
elif optinfo:
print "\n dI/df/I offdiagonals per error source:"
keys= sorted( hinfos.keys() )
#keys.remove( "01stat" )
keys= [ key for key in keys if not "stat" in key ]
for key in keys:
print "{0:>10s}:".format( stripLeadingDigits( key ) )
infom= hinfos[key]
print " ",
for name in names[1:]:
print "{0:>7s}".format( name ),
print
for i in range( nval-1 ):
for j in range( nval ):
if j == 0 and i < nval-1:
print "{0:>7s}".format( names[i] ),
elif j > i:
print "{0:7.4f}".format( infom[i,j] ),
else:
print " ",
print
print
return