def __init__(self, plotter, item):
import yaml
super().__init__(plotter, item)
if 'constraints' not in item:
raise KeyError('no constraints specified')
# extract information
self.names = item['constraints']
self.rotation = 'vertical' if 'rotation' not in item else item['rotation']
self.constraints = []
if type(self.names) == str:
self.names = [self.names]
constraints = eos.Constraints()
for name in self.names:
entry = constraints[name]
if not entry:
raise ValueError('unknown constraint {}'.format(name))
constraint = yaml.load(entry.serialize(), Loader=yaml.SafeLoader)
self.constraints.append(constraint)
def plot_constraint(self, item):
import yaml
constraints = eos.Constraints()
if 'constraints' not in item:
raise KeyError('no constraints specified')
names = item['constraints']
if type(names) == str:
names = [names]
for name in names:
entry = constraints[name]
if not entry:
raise ValueError('unknown constraint {}'.format(name))
constraint = yaml.load(entry.serialize())
xvalues = None
xerrors = None
yvalues = None
yerrors = None
if constraint['type'] == 'Gaussian':
kinematics = constraint['kinematics']
width = 1
if item['variable'] in kinematics:
xvalues = [kinematics[item['variable']]]
xerrors = None
elif (item['variable'] + '_min'
in kinematics) and (item['variable'] + '_max'
in kinematics):
xvalues = [(kinematics[item['variable'] + '_max'] +
kinematics[item['variable'] + '_min']) / 2]
xerrors = [(kinematics[item['variable'] + '_max'] -
kinematics[item['variable'] + '_min']) / 2]
if item['rescale-by-width']:
width = (kinematics[item['variable'] + '_max'] -
kinematics[item['variable'] + '_min'])
yvalues = [constraint['mean'] / width]
sigma_hi = np.sqrt(
float(constraint['sigma-stat']['hi'])**2 +
float(constraint['sigma-sys']['hi'])**2) / width
sigma_lo = np.sqrt(
float(constraint['sigma-stat']['lo'])**2 +
float(constraint['sigma-sys']['lo'])**2) / width
yerrors = [[sigma_hi, sigma_lo]]
elif constraint['type'] == 'MultivariateGaussian(Covariance)':
if 'observable' not in item:
raise KeyError(
'observable needs to be specified for MultivariateGaussian(Covariance) constraints'
)
dim = constraint['dim']
covariance = np.array(constraint['covariance'])
observables = constraint['observables']
means = constraint['means']
kinematics = constraint['kinematics']
xvalues = []
xerrors = []
yvalues = []
yerrors = []
for i in range(0, dim):
width = 1
if not observables[i] == item['observable']:
continue
_kinematics = kinematics[i]
if item['variable'] in _kinematics:
xvalues.append(_kinematics[item['variable']])
xerrors.append(0)
elif (item['variable'] + '_min'
in _kinematics) and (item['variable'] + '_max'
in _kinematics):
xvalues.append(
(_kinematics[item['variable'] + '_max'] +
_kinematics[item['variable'] + '_min']) / 2)
xerrors.append(
(_kinematics[item['variable'] + '_max'] -
_kinematics[item['variable'] + '_min']) / 2)
if item['rescale-by-width']:
width = (_kinematics[item['variable'] + '_max'] -
_kinematics[item['variable'] + '_min'])
yvalues.append(means[i] / width)
yerrors.append(np.sqrt(covariance[i, i]) / width)
elif constraint['type'] == 'MultivariateGaussian':
if 'observable' not in item:
raise KeyError(
'observable needs to be specified for MultivariateGaussian constraints'
)
dim = constraint['dim']
sigma_stat_hi = np.array(constraint['sigma-stat-hi'])
sigma_stat_lo = np.array(constraint['sigma-stat-lo'])
sigma_sys = np.array(constraint['sigma-sys'])
sigma = np.sqrt(
np.power(sigma_sys, 2) +
0.25 * np.power(sigma_stat_hi + sigma_stat_lo, 2))
observables = constraint['observables']
means = constraint['means']
kinematics = constraint['kinematics']
xvalues = []
xerrors = []
yvalues = []
yerrors = []
for i in range(0, dim):
width = 1
if not observables[i] == item['observable']:
continue
_kinematics = kinematics[i]
if item['variable'] in _kinematics:
xvalues.append(_kinematics[item['variable']])
xerrors.append(0)
elif (item['variable'] + '_min'
in _kinematics) and (item['variable'] + '_max'
in _kinematics):
xvalues.append(
(_kinematics[item['variable'] + '_max'] +
_kinematics[item['variable'] + '_min']) / 2)
xerrors.append(
(_kinematics[item['variable'] + '_max'] -
_kinematics[item['variable'] + '_min']) / 2)
if item['rescale-by-width']:
width = (_kinematics[item['variable'] + '_max'] -
_kinematics[item['variable'] + '_min'])
yvalues.append(means[i] / width)
yerrors.append(sigma[i] / width)
else:
raise ValueError('type of constraint presently not supported')
xvalues = np.array(xvalues)
if xerrors:
xerrors = np.array(xerrors)
yvalues = np.array(yvalues)
yerrors = np.array(yerrors)
color = item['color'] if 'color' in item else 'black'
label = item['label'] if 'label' in item else None
plt.errorbar(x=xvalues,
y=yvalues,
xerr=xerrors,
yerr=yerrors.T,
color=color,
elinewidth=1.0,
fmt='_',
linestyle='none',
label=label)
#!/usr/bin/python3
# vim: set sw=4 sts=4 et tw=120 :
import eos
import re
import yaml
constraints = eos.Constraints()
def latex_to_rst(s):
return (re.sub(r'\$([^\$]*)\$', r':math:`\1`', s))
page_title = 'List of Constraints'
print('#' * len(page_title))
print(page_title)
print('#' * len(page_title))
print('\n')
print(
'The following is the full list of likelihood constraints (both experimental and theoretical) included in EOS v{}.\n\n'
.format(eos.__version__))
print('\n')
print('.. list-table::')
print(' :widths: 50,50')
print(' :header-rows: 1')
print('')
print(' * - Qualified Name')
print(' - Observables')
for qn, entry in constraints:
print(' * - ``{qn}``'.format(qn=qn))
def plot(self):
import yaml
constraints = eos.Constraints()
for name in self.names:
entry = constraints[name]
if not entry:
raise ValueError('unknown constraint {}'.format(name))
constraint = yaml.load(entry.serialize(), Loader=yaml.SafeLoader)
xvalues = None
xerrors = None
yvalues = None
yerrors = None
if constraint['type'] == 'Gaussian':
kinematics = constraint['kinematics']
width = 1
if self.variable in kinematics:
xvalues = [kinematics[self.variable]]
xerrors = None
elif (self.variable + '_min' in kinematics) and (self.variable + '_max' in kinematics):
xvalues = [(kinematics[self.variable + '_max'] + kinematics[self.variable + '_min']) / 2]
xerrors = [(kinematics[self.variable + '_max'] - kinematics[self.variable + '_min']) / 2]
if self.rescale_by_width:
width = (kinematics[self.variable + '_max'] - kinematics[self.variable + '_min'])
yvalues = [float(constraint['mean']) / width]
sigma_hi = np.sqrt(float(constraint['sigma-stat']['hi'])**2 + float(constraint['sigma-sys']['hi'])**2) / width
sigma_lo = np.sqrt(float(constraint['sigma-stat']['lo'])**2 + float(constraint['sigma-sys']['lo'])**2) / width
yerrors = [[sigma_hi, sigma_lo]]
elif constraint['type'] == 'MultivariateGaussian(Covariance)':
if not self.observable:
raise KeyError('observable needs to be specified for MultivariateGaussian(Covariance) constraints')
dim = constraint['dim']
covariance = np.array(constraint['covariance'])
observables = constraint['observables']
means = constraint['means']
kinematics = constraint['kinematics']
xvalues = []
xerrors = []
yvalues = []
yerrors = []
for i in range(0, dim):
width = 1
if not observables[i] == self.observable:
continue
_kinematics = kinematics[i]
if self.variable in _kinematics:
xvalues.append(_kinematics[self.variable])
xerrors = None
elif (self.variable + '_min' in _kinematics) and (self.variable + '_max' in _kinematics):
xvalues.append((_kinematics[self.variable + '_max'] + _kinematics[self.variable + '_min']) / 2)
xerrors.append((_kinematics[self.variable + '_max'] - _kinematics[self.variable + '_min']) / 2)
if self.rescale_by_width:
width = (_kinematics[self.variable + '_max'] - _kinematics[self.variable + '_min'])
yvalues.append(means[i] / width)
yerrors.append(np.sqrt(covariance[i, i]) / width)
elif constraint['type'] == 'MultivariateGaussian':
if not self.observable:
raise KeyError('observable needs to be specified for MultivariateGaussian constraints')
dim = constraint['dim']
sigma_stat_hi = np.array(constraint['sigma-stat-hi'])
sigma_stat_lo = np.array(constraint['sigma-stat-lo'])
sigma_sys = np.array(constraint['sigma-sys'])
sigma = np.sqrt(np.power(sigma_sys, 2) + 0.25 * np.power(sigma_stat_hi + sigma_stat_lo, 2))
observables = constraint['observables']
means = constraint['means']
kinematics = constraint['kinematics']
xvalues = []
xerrors = []
yvalues = []
yerrors = []
for i in range(0, dim):
width = 1
if not observables[i] == self.observable:
continue
_kinematics = kinematics[i]
if self.variable in _kinematics:
xvalues.append(_kinematics[self.variable])
xerrors = None
elif (self.variable + '_min' in _kinematics) and (self.variable + '_max' in _kinematics):
xvalues.append((_kinematics[self.variable + '_max'] + _kinematics[self.variable + '_min']) / 2)
xerrors.append((_kinematics[self.variable + '_max'] - _kinematics[self.variable + '_min']) / 2)
if self.rescale_by_width:
width = (_kinematics[self.variable + '_max'] - _kinematics[self.variable + '_min'])
yvalues.append(means[i] / width)
yerrors.append(sigma[i] / width)
else:
raise ValueError('type of constraint presently not supported')
xvalues = np.array(xvalues)
if xerrors:
xerrors = np.array(xerrors)
yvalues = np.array(yvalues)
yerrors = np.array(yerrors)
plt.errorbar(x=xvalues, y=yvalues, xerr=xerrors, yerr=yerrors.T,
color=self.color, elinewidth=1.0, fmt='_', linestyle='none', label=self.label)
# disable the label for subsequent plots
self.label = None