from UQpy.inference.information_criteria.baseclass.InformationCriterion import InformationCriterion
from UQpy.inference import MLE
from UQpy.utilities.ValidationTypes import NumpyFloatArray
[docs]class AICc(InformationCriterion):
def minimize_criterion(self,
data: NumpyFloatArray,
parameter_estimator: MLE,
return_penalty: bool = False):
inference_model = parameter_estimator.inference_model
max_log_like = parameter_estimator.max_log_like
n_parameters = inference_model.n_parameters
n_data = len(data)
penalty_term = self._calculate_penalty_term(n_data, n_parameters)
if return_penalty:
return -2 * max_log_like + penalty_term, penalty_term
return -2 * max_log_like + penalty_term
def _calculate_penalty_term(self, n_data, n_parameters):
return 2 * n_parameters + (2 * n_parameters ** 2 + 2 * n_parameters) / (n_data - n_parameters - 1)