models.py 文件源码

def _estimate_bootstrapped_params(self):
        """
        Compute the bootstrapped parameters for:
            * The path from the predictors to Y (computed using OLS/Logit, depending on the nature of Y)
            * The path(s) from the mediator(s) to Y (computed using OLS)
        :return: A tuple of (true_betas_y, true_betas_m)
            * true_betas_y is a matrix of size n_boots x n_params_y
            * true_betas_m is a list of matrices of size n_boots x n_params_y
        """
        n_boots = self._options["boot"]
        seed = self._options["seed"]
        boot_betas_y = np.empty((n_boots, len(self._exog_terms_y)))
        boot_betas_m = np.empty((self._n_meds, n_boots, len(self._exog_terms_m)))
        n_fail_samples = 0
        boot_ind = 0
        sampler = bootstrap_sampler(self._n_obs, seed)
        while boot_ind < n_boots:
            ind = next(sampler)
            data_boot = self._data[ind, :]
            y_e = data_boot[:, self._ind_y]
            y_x = data_boot[:, self._exog_inds_y]
            try:
                y_b = self._compute_betas_y(y_e, y_x)
                m_x = data_boot[:, self._exog_inds_m]
                boot_betas_y[boot_ind] = y_b
                for j, m_ind in enumerate(self._inds_m):
                    m_e = data_boot[:, m_ind]
                    m_b = self._compute_betas_m(m_e, m_x)
                    boot_betas_m[j][boot_ind] = m_b
                boot_ind += 1
            except LinAlgError: # Hessian (Logit) or X'X (OLS) cannot be inverted
                n_fail_samples += 1

        return boot_betas_y, boot_betas_m, n_fail_samples