def _build_results(self):
"""
build the results after a fit is performed
:returns:
:rtype:
"""
# set the median fit
self.restore_median_fit()
# Find maximum of the log posterior
idx = self._log_probability_values.argmax()
# Get parameter values at the maximum
approximate_MAP_point = self._raw_samples[idx, :]
# Sets the values of the parameters to their MAP values
for i, parameter in enumerate(self._free_parameters):
self._free_parameters[parameter].value = approximate_MAP_point[i]
# Get the value of the posterior for each dataset at the MAP
log_posteriors = collections.OrderedDict()
log_prior = self._log_prior(approximate_MAP_point)
# keep track of the total number of data points
# and the total posterior
total_n_data_points = 0
total_log_posterior = 0
for dataset in list(self._data_list.values()):
log_posterior = dataset.get_log_like() + log_prior
log_posteriors[dataset.name] = log_posterior
total_n_data_points += dataset.get_number_of_data_points()
total_log_posterior += log_posterior
# compute the statistical measures
statistical_measures = collections.OrderedDict()
# compute the point estimates
statistical_measures["AIC"] = aic(total_log_posterior,
len(self._free_parameters),
total_n_data_points)
statistical_measures["BIC"] = bic(total_log_posterior,
len(self._free_parameters),
total_n_data_points)
this_dic, pdic = dic(self)
# compute the posterior estimates
statistical_measures["DIC"] = this_dic
statistical_measures["PDIC"] = pdic
if self._marginal_likelihood is not None:
statistical_measures["log(Z)"] = self._marginal_likelihood
# TODO: add WAIC
# Instance the result
self._results = BayesianResults(
self._likelihood_model,
self._raw_samples,
log_posteriors,
statistical_measures=statistical_measures,
log_probabilty=self._log_like_values)
def _build_results(self):
# Find maximum of the log posterior
idx = self._log_probability_values.argmax()
# Get parameter values at the maximum
approximate_MAP_point = self._raw_samples[idx, :]
# Sets the values of the parameters to their MAP values
for i, parameter in enumerate(self._free_parameters):
self._free_parameters[parameter].value = approximate_MAP_point[i]
# Get the value of the posterior for each dataset at the MAP
log_posteriors = collections.OrderedDict()
log_prior = self._log_prior(approximate_MAP_point)
# keep track of the total number of data points
# and the total posterior
total_n_data_points = 0
total_log_posterior = 0
for dataset in self._data_list.values():
log_posterior = dataset.get_log_like() + log_prior
log_posteriors[dataset.name] = log_posterior
total_n_data_points += dataset.get_number_of_data_points()
total_log_posterior += log_posterior
# compute the statistical measures
statistical_measures = collections.OrderedDict()
# compute the point estimates
statistical_measures['AIC'] = aic(total_log_posterior,len(self._free_parameters),total_n_data_points)
statistical_measures['BIC'] = bic(total_log_posterior,len(self._free_parameters),total_n_data_points)
this_dic, pdic = dic(self)
# compute the posterior estimates
statistical_measures['DIC'] = this_dic
statistical_measures['PDIC'] = pdic
if self._marginal_likelihood is not None:
statistical_measures['log(Z)'] = self._marginal_likelihood
#TODO: add WAIC
# Instance the result
self._results = BayesianResults(self._likelihood_model, self._raw_samples, log_posteriors,statistical_measures=statistical_measures)