def exp_measurements(self):
"""Return a list of four probability distributions, corresponding
to the experimental measurements of the four input observables."""
return {
obs: flavio.combine_measurements(obs)
for obs in self.observables
}
def test_exp_combo(self):
o = Observable('test_obs')
o.arguments = ['x']
m = Measurement('test_obs measurement 1')
m.add_constraint([('test_obs', 1)], MultivariateNormalDistribution([1, 2], np.eye(2)))
# error: no measurement
with self.assertRaises(ValueError):
flavio.combine_measurements('test_obs', x=1, include_measurements=['bla'])
m.add_constraint([('test_obs', 1)], NormalDistribution(2, 3))
combo = flavio.combine_measurements('test_obs', x=1)
self.assertEqual(combo.central_value, 2)
self.assertEqual(combo.standard_deviation, 3)
m2 = Measurement('test_obs measurement 2')
m2.add_constraint([('test_obs', 1)], NormalDistribution(3, 3))
combo = flavio.combine_measurements('test_obs', x=1)
self.assertAlmostEqual(combo.central_value, 2.5)
self.assertAlmostEqual(combo.standard_deviation, sqrt(9 / 2))
Observable.del_instance('test_obs')
def _obstable_tree(self):
if not self._obstable_tree_cache:
info = tree() # nested dict
pull_dof = 1
llh = self.likelihood
for flh_name, flh in llh.fast_likelihoods.items():
# loop over fast likelihoods: they only have a single "measurement"
m = flh.pseudo_measurement
ml = flh.full_measurement_likelihood
pred = ml.get_predictions_par(llh.par_dict, self.w)
sm_cov = flh.sm_covariance.get(force=False)
_, exp_cov = flh.exp_covariance.get(force=False)
inspire_dict = self._get_inspire_dict(flh.observables, ml)
for i, obs in enumerate(flh.observables):
info[obs]['lh_name'] = flh_name
info[obs]['name'] = obs if isinstance(obs, str) else obs[0]
info[obs]['theory'] = pred[obs]
info[obs]['th. unc.'] = np.sqrt(sm_cov[i, i])
info[obs]['experiment'] = m.get_central(obs)
info[obs]['exp. unc.'] = np.sqrt(exp_cov[i, i])
info[obs]['exp. PDF'] = NormalDistribution(
m.get_central(obs), np.sqrt(exp_cov[i, i]))
info[obs]['inspire'] = sorted(set(inspire_dict[obs]))
ll_central = m.get_logprobability_single(
obs, m.get_central(obs))
ll = m.get_logprobability_single(obs, pred[obs])
# DeltaChi2 is -2*DeltaLogLikelihood
info[obs]['pull'] = pull(-2 * (ll - ll_central),
dof=pull_dof)
for lh_name, lh in llh.likelihoods.items():
# loop over "normal" likelihoods
ml = lh.measurement_likelihood
pred = ml.get_predictions_par(llh.par_dict, self.w)
inspire_dict = self._get_inspire_dict(lh.observables, ml)
for i, obs in enumerate(lh.observables):
obs_dict = flavio.Observable.argument_format(obs, 'dict')
obs_name = obs_dict.pop('name')
with warnings.catch_warnings():
warnings.simplefilter("ignore")
p_comb = flavio.combine_measurements(
obs_name,
include_measurements=ml.get_measurements,
**obs_dict)
info[obs]['experiment'] = p_comb.central_value
info[obs]['exp. unc.'] = max(p_comb.error_left,
p_comb.error_right)
info[obs]['exp. PDF'] = p_comb
info[obs]['inspire'] = sorted(set(inspire_dict[obs]))
info[obs]['theory'] = pred[obs]
info[obs]['th. unc.'] = 0
info[obs]['lh_name'] = lh_name
info[obs]['name'] = obs if isinstance(obs, str) else obs[0]
ll = p_comb.logpdf([pred[obs]]) - p_comb.logpdf(
[p_comb.central_value])
info[obs]['pull'] = pull(-2 * ll, dof=pull_dof)
self._obstable_tree_cache = info
return self._obstable_tree_cache
def test_ae_SM(self):
self.assertAlmostEqual(flavio.sm_prediction('a_e') / ae_SM, 1)
pd = flavio.combine_measurements('a_e')
ae_exp = pd.central_value
ae_err_exp = pd.error_left
np.random.seed(17)
ae_err_sm = flavio.sm_uncertainty('a_e')
# check that there is a -2.3 sigma tension, see 1804.07409 p. 13
self.assertAlmostEqual((ae_exp - ae_SM) / sqrt(ae_err_sm**2 + ae_err_exp**2), -2.3, delta=0.5)
def obstable_sm(self):
self._check_sm_cov_loaded()
if self._obstable_sm is None:
info = tree() # nested dict
for flh_name, flh in self.fast_likelihoods.items():
# loop over fast likelihoods: they only have a single "measurement"
m = flh.pseudo_measurement
ml = flh.full_measurement_likelihood
pred_sm = ml.get_predictions_par(self.par_dict_sm,
flavio.WilsonCoefficients())
sm_cov = flh.sm_covariance.get(force=False)
_, exp_cov = flh.exp_covariance.get(force=False)
inspire_dict = self._get_inspire_dict(flh.observables, ml)
for i, obs in enumerate(flh.observables):
info[obs]['lh_name'] = flh_name
info[obs]['name'] = obs if isinstance(obs, str) else obs[0]
info[obs]['th. unc.'] = np.sqrt(sm_cov[i, i])
info[obs]['experiment'] = m.get_central(obs)
info[obs]['exp. unc.'] = np.sqrt(exp_cov[i, i])
info[obs]['exp. PDF'] = NormalDistribution(
m.get_central(obs), np.sqrt(exp_cov[i, i]))
info[obs]['inspire'] = sorted(set(inspire_dict[obs]))
info[obs]['ll_sm'] = m.get_logprobability_single(
obs, pred_sm[obs])
info[obs]['ll_central'] = m.get_logprobability_single(
obs, m.get_central(obs))
for lh_name, lh in self.likelihoods.items():
# loop over "normal" likelihoods
ml = lh.measurement_likelihood
pred_sm = ml.get_predictions_par(self.par_dict_sm,
flavio.WilsonCoefficients())
inspire_dict = self._get_inspire_dict(lh.observables, ml)
for i, obs in enumerate(lh.observables):
obs_dict = flavio.Observable.argument_format(obs, 'dict')
obs_name = obs_dict.pop('name')
with warnings.catch_warnings():
warnings.simplefilter("ignore")
p_comb = flavio.combine_measurements(
obs_name,
include_measurements=ml.get_measurements,
**obs_dict)
info[obs]['experiment'] = p_comb.central_value
info[obs]['exp. unc.'] = max(p_comb.error_left,
p_comb.error_right)
info[obs]['exp. PDF'] = p_comb
info[obs]['inspire'] = sorted(set(inspire_dict[obs]))
info[obs]['th. unc.'] = 0
info[obs]['lh_name'] = lh_name
info[obs]['name'] = obs if isinstance(obs, str) else obs[0]
info[obs]['ll_sm'] = p_comb.logpdf([pred_sm[obs]])
if info[obs]['ll_sm'] == -np.inf:
info[obs]['ll_sm'] = -1e100
info[obs]['ll_central'] = p_comb.logpdf(
[p_comb.central_value])
self._obstable_sm = info
return self._obstable_sm
def test_ae_SM(self):
self.assertAlmostEqual(flavio.sm_prediction('a_e') / ae_SM, 1)
pd = flavio.combine_measurements('a_e')
ae_exp = pd.central_value
ae_err_exp = pd.error_left
np.random.seed(17)
ae_err_sm = flavio.sm_uncertainty('a_e')
# check that there is a -2.3 sigma tension, see 1804.07409 p. 13
self.assertAlmostEqual(
(ae_exp - ae_SM) / sqrt(ae_err_sm**2 + ae_err_exp**2),
-2.3,
delta=0.5)
def test_exp(self):
self.assertAlmostEqual(flavio.combine_measurements('tau_Bc').central_value
/ par['tau_Bc'], 1,
delta=0.001)
