def _build_res(self, res: OptimizeResult):
res.log_likelihood = -res.fun
res.predictive_power = (self.predict_permuted(res.x) *
self.is_chosen).sum() / len(self.df)
res.aic = res.log_likelihood - self._n_free_params()
res.bic = res.log_likelihood - self._n_free_params() * np.log(
len(self.df)) / 2
res.π_L = self.π_L
res.β = self.β
res.b = self.b
def load_model(self, Model: Type[models.Model],
res: OptimizeResult) -> models.Model:
if len(res.b) > 1:
res.b = np.array([res.b[-1] - res.b[-2]])
return super().load_model(Model, res)
