diff --git a/econml/dml/dml.py b/econml/dml/dml.py index 5cc84d61d..efc9931e3 100644 --- a/econml/dml/dml.py +++ b/econml/dml/dml.py @@ -653,6 +653,18 @@ class SparseLinearDML(DebiasedLassoCateEstimatorMixin, DML): CATE L1 regularization applied through the debiased lasso in the final model. 'auto' corresponds to a CV form of the :class:`MultiOutputDebiasedLasso`. + n_alphas : int, optional, default 100 + How many alphas to try if alpha='auto' + + alpha_cov : string | float, optional, default 'auto' + The regularization alpha that is used when constructing the pseudo inverse of + the covariance matrix Theta used to for correcting the final state lasso coefficient + in the debiased lasso. Each such regression corresponds to the regression of one feature + on the remainder of the features. + + n_alphas_cov : int, optional, default 10 + How many alpha_cov to try if alpha_cov='auto'. + max_iter : int, optional, default=1000 The maximum number of iterations in the Debiased Lasso @@ -707,8 +719,12 @@ class SparseLinearDML(DebiasedLassoCateEstimatorMixin, DML): def __init__(self, model_y='auto', model_t='auto', alpha='auto', + n_alphas=100, + alpha_cov='auto', + n_alphas_cov=10, max_iter=1000, tol=1e-4, + n_jobs=None, featurizer=None, fit_cate_intercept=True, linear_first_stages=True, @@ -718,9 +734,13 @@ def __init__(self, random_state=None): model_final = MultiOutputDebiasedLasso( alpha=alpha, + n_alphas=n_alphas, + alpha_cov=alpha_cov, + n_alphas_cov=n_alphas_cov, fit_intercept=False, max_iter=max_iter, tol=tol, + n_jobs=n_jobs, random_state=random_state) super().__init__(model_y=model_y, model_t=model_t, diff --git a/econml/drlearner.py b/econml/drlearner.py index 87d58de2f..8f220ad4b 100644 --- a/econml/drlearner.py +++ b/econml/drlearner.py @@ -853,6 +853,18 @@ class SparseLinearDRLearner(DebiasedLassoCateEstimatorDiscreteMixin, DRLearner): CATE L1 regularization applied through the debiased lasso in the final model. 'auto' corresponds to a CV form of the :class:`DebiasedLasso`. + n_alphas : int, optional, default 100 + How many alphas to try if alpha='auto' + + alpha_cov : string | float, optional, default 'auto' + The regularization alpha that is used when constructing the pseudo inverse of + the covariance matrix Theta used to for correcting the final state lasso coefficient + in the debiased lasso. Each such regression corresponds to the regression of one feature + on the remainder of the features. + + n_alphas_cov : int, optional, default 10 + How many alpha_cov to try if alpha_cov='auto'. + max_iter : int, optional, default 1000 The maximum number of iterations in the Debiased Lasso @@ -910,17 +922,17 @@ class SparseLinearDRLearner(DebiasedLassoCateEstimatorDiscreteMixin, DRLearner): est.fit(y, T, X=X, W=None) >>> est.effect(X[:3]) - array([ 0.418400..., 0.306400..., -0.130733...]) + array([ 0.41..., 0.31..., -0.12...]) >>> est.effect_interval(X[:3]) - (array([ 0.056783..., -0.206438..., -0.739296...]), array([0.780017..., 0.819239..., 0.477828...])) + (array([ 0.04..., -0.19..., -0.73...]), array([0.77..., 0.82..., 0.47...])) >>> est.coef_(T=1) - array([0.449779..., 0.004807..., 0.061954...]) + array([ 0.45..., -0.00..., 0.06...]) >>> est.coef__interval(T=1) - (array([ 0.242194... , -0.190825..., -0.139646...]), array([0.657365..., 0.200440..., 0.263556...])) + (array([ 0.24... , -0.19..., -0.13...]), array([0.65..., 0.19..., 0.26...])) >>> est.intercept_(T=1) - 0.88436847... + 0.88... >>> est.intercept__interval(T=1) - (0.68683788..., 1.08189907...) + (0.68..., 1.08...) Attributes ---------- @@ -942,17 +954,25 @@ def __init__(self, featurizer=None, fit_cate_intercept=True, alpha='auto', + n_alphas=100, + alpha_cov='auto', + n_alphas_cov=10, max_iter=1000, tol=1e-4, min_propensity=1e-6, categories='auto', - n_splits=2, random_state=None): + n_splits=2, + random_state=None): self.fit_cate_intercept = fit_cate_intercept model_final = DebiasedLasso( alpha=alpha, + n_alphas=n_alphas, + alpha_cov=alpha_cov, + n_alphas_cov=n_alphas_cov, fit_intercept=fit_cate_intercept, max_iter=max_iter, - tol=tol) + tol=tol, + random_state=random_state) super().__init__(model_propensity=model_propensity, model_regression=model_regression, model_final=model_final, diff --git a/econml/sklearn_extensions/linear_model.py b/econml/sklearn_extensions/linear_model.py index 4b20d3ebb..6fe6311ff 100644 --- a/econml/sklearn_extensions/linear_model.py +++ b/econml/sklearn_extensions/linear_model.py @@ -35,6 +35,7 @@ from statsmodels.tools.tools import add_constant from statsmodels.api import RLM import statsmodels +from joblib import Parallel, delayed def _weighted_check_cv(cv=5, y=None, classifier=False): @@ -539,6 +540,34 @@ def fit(self, X, y, sample_weight=None): return self +def _get_theta_coefs_and_tau_sq(i, X, sample_weight, alpha_cov, n_alphas_cov, max_iter, tol, random_state): + n_samples, n_features = X.shape + y = X[:, i] + X_reduced = X[:, list(range(i)) + list(range(i + 1, n_features))] + # Call weighted lasso on reduced design matrix + if alpha_cov == 'auto': + local_wlasso = WeightedLassoCV(cv=3, n_alphas=n_alphas_cov, + fit_intercept=False, + max_iter=max_iter, + tol=tol, n_jobs=1, + random_state=random_state) + else: + local_wlasso = WeightedLasso(alpha=alpha_cov, + fit_intercept=False, + max_iter=max_iter, + tol=tol, + random_state=random_state) + local_wlasso.fit(X_reduced, y, sample_weight=sample_weight) + coefs = local_wlasso.coef_ + # Weighted tau + if sample_weight is not None: + y_weighted = y * sample_weight / np.sum(sample_weight) + else: + y_weighted = y / n_samples + tausq = np.dot(y - local_wlasso.predict(X_reduced), y_weighted) + return coefs, tausq + + class DebiasedLasso(WeightedLasso): """Debiased Lasso model. @@ -555,6 +584,18 @@ class DebiasedLasso(WeightedLasso): reasons, using ``alpha = 0`` with the ``Lasso`` object is not advised. Given this, you should use the :class:`.LinearRegression` object. + n_alphas : int, optional, default 100 + How many alphas to try if alpha='auto' + + alpha_cov : string | float, optional, default 'auto' + The regularization alpha that is used when constructing the pseudo inverse of + the covariance matrix Theta used to for correcting the lasso coefficient. Each + such regression corresponds to the regression of one feature on the remainder + of the features. + + n_alphas_cov : int, optional, default 10 + How many alpha_cov to try if alpha_cov='auto'. + fit_intercept : boolean, optional, default True Whether to calculate the intercept for this model. If set to False, no intercept will be used in calculations @@ -597,6 +638,9 @@ class DebiasedLasso(WeightedLasso): (setting to 'random') often leads to significantly faster convergence especially when tol is higher than 1e-4. + n_jobs : int or None, default None + How many jobs to use whenever parallelism is invoked + Attributes ---------- coef_ : array, shape (n_features,) @@ -620,10 +664,14 @@ class DebiasedLasso(WeightedLasso): """ - def __init__(self, alpha='auto', fit_intercept=True, - precompute=False, copy_X=True, max_iter=1000, + def __init__(self, alpha='auto', n_alphas=100, alpha_cov='auto', n_alphas_cov=10, + fit_intercept=True, precompute=False, copy_X=True, max_iter=1000, tol=1e-4, warm_start=False, - random_state=None, selection='cyclic'): + random_state=None, selection='cyclic', n_jobs=None): + self.n_jobs = n_jobs + self.n_alphas = n_alphas + self.alpha_cov = alpha_cov + self.n_alphas_cov = n_alphas_cov super().__init__( alpha=alpha, fit_intercept=fit_intercept, precompute=precompute, copy_X=copy_X, @@ -747,18 +795,8 @@ def predict_interval(self, X, alpha=0.1): lower = alpha / 2 upper = 1 - alpha / 2 y_pred = self.predict(X) - y_lower = np.empty(y_pred.shape) - y_upper = np.empty(y_pred.shape) - # Note that in the case of no intercept, X_offset is 0 - if self.fit_intercept: - X = X - self._X_offset - # Calculate the variance of the predictions - var_pred = np.sum(np.matmul(X, self._coef_variance) * X, axis=1) - if self.fit_intercept: - var_pred += self._mean_error_variance - # Calculate prediction confidence intervals - sd_pred = np.sqrt(var_pred) + sd_pred = self.prediction_stderr(X) y_lower = y_pred + \ np.apply_along_axis(lambda s: norm.ppf( lower, scale=s), 0, sd_pred) @@ -810,7 +848,7 @@ def intercept__interval(self, alpha=0.1): def _get_coef_correction(self, X, y, y_pred, sample_weight, theta_hat): # Assumes flattened y - n_samples, n_features = X.shape + n_samples, _ = X.shape y_res = np.ndarray.flatten(y) - y_pred # Compute weighted residuals if sample_weight is not None: @@ -818,12 +856,17 @@ def _get_coef_correction(self, X, y, y_pred, sample_weight, theta_hat): else: y_res_scaled = y_res / n_samples delta_coef = np.matmul( - np.matmul(theta_hat, X.T), y_res_scaled) + theta_hat, np.matmul(X.T, y_res_scaled)) return delta_coef def _get_optimal_alpha(self, X, y, sample_weight): # To be done once per target. Assumes y can be flattened. - cv_estimator = WeightedLassoCV(cv=5, fit_intercept=self.fit_intercept) + cv_estimator = WeightedLassoCV(cv=5, n_alphas=self.n_alphas, fit_intercept=self.fit_intercept, + precompute=self.precompute, copy_X=True, + max_iter=self.max_iter, tol=self.tol, + random_state=self.random_state, + selection=self.selection, + n_jobs=self.n_jobs) cv_estimator.fit(X, y.flatten(), sample_weight=sample_weight) return cv_estimator.alpha_ @@ -835,27 +878,15 @@ def _get_theta_hat(self, X, sample_weight): C_hat = np.ones((1, 1)) tausq = (X.T @ X / n_samples).flatten() return np.diag(1 / tausq) @ C_hat - coefs = np.empty((n_features, n_features - 1)) - tausq = np.empty(n_features) # Compute Lasso coefficients for the columns of the design matrix - for i in range(n_features): - y = X[:, i] - X_reduced = X[:, list(range(i)) + list(range(i + 1, n_features))] - # Call weighted lasso on reduced design matrix - # Inherit some parameters from the parent - local_wlasso = WeightedLasso( - alpha=self.alpha, - fit_intercept=False, - max_iter=self.max_iter, - tol=self.tol - ).fit(X_reduced, y, sample_weight=sample_weight) - coefs[i] = local_wlasso.coef_ - # Weighted tau - if sample_weight is not None: - y_weighted = y * sample_weight / np.sum(sample_weight) - else: - y_weighted = y / n_samples - tausq[i] = np.dot(y - local_wlasso.predict(X_reduced), y_weighted) + results = Parallel(n_jobs=self.n_jobs)( + delayed(_get_theta_coefs_and_tau_sq)(i, X, sample_weight, + self.alpha_cov, self.n_alphas_cov, + self.max_iter, self.tol, self.random_state) + for i in range(n_features)) + coefs, tausq = zip(*results) + coefs = np.array(coefs) + tausq = np.array(tausq) # Compute C_hat C_hat = np.diag(np.ones(n_features)) C_hat[0][1:] = -coefs[0] @@ -893,6 +924,18 @@ class MultiOutputDebiasedLasso(MultiOutputRegressor): reasons, using ``alpha = 0`` with the ``Lasso`` object is not advised. Given this, you should use the :class:`LinearRegression` object. + n_alphas : int, optional, default 100 + How many alphas to try if alpha='auto' + + alpha_cov : string | float, optional, default 'auto' + The regularization alpha that is used when constructing the pseudo inverse of + the covariance matrix Theta used to for correcting the lasso coefficient. Each + such regression corresponds to the regression of one feature on the remainder + of the features. + + n_alphas_cov : int, optional, default 10 + How many alpha_cov to try if alpha_cov='auto'. + fit_intercept : boolean, optional, default True Whether to calculate the intercept for this model. If set to False, no intercept will be used in calculations @@ -935,6 +978,9 @@ class MultiOutputDebiasedLasso(MultiOutputRegressor): (setting to 'random') often leads to significantly faster convergence especially when tol is higher than 1e-4. + n_jobs : int or None, default None + How many jobs to use whenever parallelism is invoked + Attributes ---------- coef_ : array, shape (n_targets, n_features) or (n_features,) @@ -954,14 +1000,17 @@ class MultiOutputDebiasedLasso(MultiOutputRegressor): """ - def __init__(self, alpha='auto', fit_intercept=True, + def __init__(self, alpha='auto', n_alphas=100, alpha_cov='auto', n_alphas_cov=10, + fit_intercept=True, precompute=False, copy_X=True, max_iter=1000, tol=1e-4, warm_start=False, random_state=None, selection='cyclic', n_jobs=None): - self.estimator = DebiasedLasso(alpha=alpha, fit_intercept=fit_intercept, + self.estimator = DebiasedLasso(alpha=alpha, n_alphas=n_alphas, alpha_cov=alpha_cov, n_alphas_cov=n_alphas_cov, + fit_intercept=fit_intercept, precompute=precompute, copy_X=copy_X, max_iter=max_iter, tol=tol, warm_start=warm_start, - random_state=random_state, selection=selection) + random_state=random_state, selection=selection, + n_jobs=n_jobs) super().__init__(estimator=self.estimator, n_jobs=n_jobs) def fit(self, X, y, sample_weight=None): diff --git a/econml/tests/test_dml.py b/econml/tests/test_dml.py index e335b4bc1..04d7e2124 100644 --- a/econml/tests/test_dml.py +++ b/econml/tests/test_dml.py @@ -967,9 +967,10 @@ def test_categories(self): dmls = [LinearDML, SparseLinearDML] for ctor in dmls: dml1 = ctor(LinearRegression(), LogisticRegression(C=1000), - fit_cate_intercept=False, discrete_treatment=True) + fit_cate_intercept=False, discrete_treatment=True, random_state=123) dml2 = ctor(LinearRegression(), LogisticRegression(C=1000), - fit_cate_intercept=False, discrete_treatment=True, categories=['c', 'b', 'a']) + fit_cate_intercept=False, discrete_treatment=True, categories=['c', 'b', 'a'], + random_state=123) # create a simple artificial setup where effect of moving from treatment # a -> b is 2, @@ -1003,9 +1004,9 @@ def test_categories(self): # but const_marginal_effect should be reordered based on the explicit cagetories cme1 = dml1.const_marginal_effect(np.ones((1, 1))).reshape(-1) cme2 = dml2.const_marginal_effect(np.ones((1, 1))).reshape(-1) - self.assertAlmostEqual(cme1[1], -cme2[1], places=4) # 1->3 in original ordering; 3->1 in new ordering + self.assertAlmostEqual(cme1[1], -cme2[1], places=3) # 1->3 in original ordering; 3->1 in new ordering # 1-> 2 in original ordering; combination of 3->1 and 3->2 - self.assertAlmostEqual(cme1[0], -cme2[1] + cme2[0], places=4) + self.assertAlmostEqual(cme1[0], -cme2[1] + cme2[0], places=3) def test_groups(self): groups = [1, 2, 3, 4, 5, 6] * 10 diff --git a/notebooks/Doubly Robust Learner and Interpretability.ipynb b/notebooks/Doubly Robust Learner and Interpretability.ipynb index 3d281411e..bda63cbd7 100644 --- a/notebooks/Doubly Robust Learner and Interpretability.ipynb +++ b/notebooks/Doubly Robust Learner and Interpretability.ipynb @@ -86,7 +86,7 @@ { "data": { "text/plain": [ - "" + "" ] }, "execution_count": 5, @@ -514,7 +514,7 @@ { "data": { "text/plain": [ - "" + "" ] }, "execution_count": 14, @@ -763,7 +763,7 @@ "

Note: The stderr_mean is a conservative upper bound." ], "text/plain": [ - "" + "" ] }, "execution_count": 17, @@ -791,7 +791,7 @@ { "data": { "text/plain": [ - "" + "" ] }, "execution_count": 18, @@ -828,7 +828,7 @@ "outputs": [ { "data": { - "image/png": "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\n", + "image/png": "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\n", "text/plain": [ "
" ] @@ -871,7 +871,7 @@ "outputs": [ { "data": { - "image/png": "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\n", + "image/png": "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\n", "text/plain": [ "
" ] @@ -962,93 +962,93 @@ " \n", " \n", " 0\n", - " 1.714\n", - " 0.484\n", - " 3.541\n", + " 1.198\n", + " 0.274\n", + " 4.370\n", " 0.000\n", - " 0.918\n", - " 2.511\n", + " 0.747\n", + " 1.648\n", " \n", " \n", " 1\n", - " -1.319\n", - " 0.268\n", - " -4.924\n", + " -1.087\n", + " 0.271\n", + " -4.009\n", " 0.000\n", - " -1.759\n", - " -0.878\n", + " -1.533\n", + " -0.641\n", " \n", " \n", " 2\n", - " 1.692\n", - " 0.890\n", - " 1.901\n", - " 0.057\n", - " 0.228\n", - " 3.156\n", + " 1.129\n", + " 0.280\n", + " 4.033\n", + " 0.000\n", + " 0.669\n", + " 1.590\n", " \n", " \n", " 3\n", - " -1.319\n", - " 0.268\n", - " -4.924\n", + " -1.087\n", + " 0.271\n", + " -4.009\n", " 0.000\n", - " -1.759\n", - " -0.878\n", + " -1.533\n", + " -0.641\n", " \n", " \n", " 4\n", - " -1.178\n", - " 0.268\n", - " -4.399\n", + " -1.095\n", + " 0.269\n", + " -4.065\n", " 0.000\n", - " -1.619\n", - " -0.738\n", + " -1.537\n", + " -0.652\n", " \n", " \n", " 5\n", - " 1.035\n", - " 0.303\n", - " 3.421\n", - " 0.001\n", - " 0.538\n", - " 1.533\n", + " 0.725\n", + " 0.291\n", + " 2.491\n", + " 0.013\n", + " 0.246\n", + " 1.204\n", " \n", " \n", " 6\n", - " 1.103\n", - " 0.476\n", - " 2.318\n", - " 0.020\n", - " 0.320\n", - " 1.886\n", + " 0.882\n", + " 0.272\n", + " 3.239\n", + " 0.001\n", + " 0.434\n", + " 1.330\n", " \n", " \n", " 7\n", - " -1.078\n", - " 0.267\n", - " -4.039\n", - " 0.000\n", - " -1.517\n", - " -0.639\n", + " -0.803\n", + " 0.273\n", + " -2.941\n", + " 0.003\n", + " -1.251\n", + " -0.354\n", " \n", " \n", " 8\n", - " 1.103\n", - " 0.476\n", - " 2.318\n", - " 0.020\n", - " 0.320\n", - " 1.886\n", + " 0.882\n", + " 0.272\n", + " 3.239\n", + " 0.001\n", + " 0.434\n", + " 1.330\n", " \n", " \n", " 9\n", - " 1.035\n", - " 0.303\n", - " 3.421\n", - " 0.001\n", - " 0.538\n", - " 1.533\n", + " 0.725\n", + " 0.291\n", + " 2.491\n", + " 0.013\n", + " 0.246\n", + " 1.204\n", " \n", " \n", "\n", @@ -1056,16 +1056,16 @@ ], "text/plain": [ " point_estimate stderr zstat pvalue ci_lower ci_upper\n", - "0 1.714 0.484 3.541 0.000 0.918 2.511\n", - "1 -1.319 0.268 -4.924 0.000 -1.759 -0.878\n", - "2 1.692 0.890 1.901 0.057 0.228 3.156\n", - "3 -1.319 0.268 -4.924 0.000 -1.759 -0.878\n", - "4 -1.178 0.268 -4.399 0.000 -1.619 -0.738\n", - "5 1.035 0.303 3.421 0.001 0.538 1.533\n", - "6 1.103 0.476 2.318 0.020 0.320 1.886\n", - "7 -1.078 0.267 -4.039 0.000 -1.517 -0.639\n", - "8 1.103 0.476 2.318 0.020 0.320 1.886\n", - "9 1.035 0.303 3.421 0.001 0.538 1.533" + "0 1.198 0.274 4.370 0.000 0.747 1.648\n", + "1 -1.087 0.271 -4.009 0.000 -1.533 -0.641\n", + "2 1.129 0.280 4.033 0.000 0.669 1.590\n", + "3 -1.087 0.271 -4.009 0.000 -1.533 -0.641\n", + "4 -1.095 0.269 -4.065 0.000 -1.537 -0.652\n", + "5 0.725 0.291 2.491 0.013 0.246 1.204\n", + "6 0.882 0.272 3.239 0.001 0.434 1.330\n", + "7 -0.803 0.273 -2.941 0.003 -1.251 -0.354\n", + "8 0.882 0.272 3.239 0.001 0.434 1.330\n", + "9 0.725 0.291 2.491 0.013 0.246 1.204" ] }, "execution_count": 21, @@ -1092,7 +1092,7 @@ " mean_point stderr_mean zstat pvalue ci_mean_lower ci_mean_upper\n", "\n", "\n", - " 0.279 0.441 0.632 0.527 -0.447 1.005 \n", + " 0.147 0.277 0.531 0.595 -0.308 0.602 \n", "\n", "\n", "\n", @@ -1101,7 +1101,7 @@ " \n", "\n", "\n", - " \n", + " \n", "\n", "
std_point pct_point_lower pct_point_upper
1.25 -1.319 1.7040.965 -1.091 1.167
\n", "\n", @@ -1110,12 +1110,12 @@ " \n", "\n", "\n", - " \n", + " \n", "\n", "
stderr_point ci_point_lower ci_point_upper
1.326 -1.55 2.1631.004 -1.359 1.388


Note: The stderr_mean is a conservative upper bound." ], "text/plain": [ - "" + "" ] }, "execution_count": 22,