def test_doubleml_exception_scores():
msg = 'Invalid score IV. Valid score IV-type or partialling out.'
with pytest.raises(ValueError, match=msg):
_ = DoubleMLPLR(dml_data, ml_g, ml_m, score='IV')
msg = 'score should be either a string or a callable. 0 was passed.'
with pytest.raises(TypeError, match=msg):
_ = DoubleMLPLR(dml_data, ml_g, ml_m, score=0)
msg = 'Invalid score IV. Valid score ATE or ATTE.'
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIRM(dml_data_irm, Lasso(), LogisticRegression(), score='IV')
msg = 'score should be either a string or a callable. 0 was passed.'
with pytest.raises(TypeError, match=msg):
_ = DoubleMLIRM(dml_data_irm, Lasso(), LogisticRegression(), score=0)
msg = 'Invalid score ATE. Valid score LATE.'
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(), LogisticRegression(), score='ATE')
msg = 'score should be either a string or a callable. 0 was passed.'
with pytest.raises(TypeError, match=msg):
_ = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(), LogisticRegression(), score=0)
msg = 'Invalid score IV. Valid score partialling out.'
with pytest.raises(ValueError, match=msg):
_ = DoubleMLPLIV(dml_data_pliv, Lasso(), Lasso(), Lasso(), score='IV')
msg = 'score should be either a string or a callable. 0 was passed.'
with pytest.raises(TypeError, match=msg):
_ = DoubleMLPLIV(dml_data_pliv, Lasso(), Lasso(), Lasso(), score=0)
def test_doubleml_exception_data():
msg = 'The data must be of DoubleMLData type.'
with pytest.raises(TypeError, match=msg):
_ = DoubleMLPLR(pd.DataFrame(), ml_g, ml_m)
msg = (
r'Incompatible data. Z1 have been set as instrumental variable\(s\). '
'To fit a partially linear IV regression model use DoubleMLPLIV instead of DoubleMLPLR.'
)
with pytest.raises(ValueError, match=msg):
_ = DoubleMLPLR(dml_data_pliv, ml_g, ml_m)
msg = (
r'Incompatible data. z have been set as instrumental variable\(s\). '
'To fit an interactive IV regression model use DoubleMLIIVM instead of DoubleMLIRM.'
)
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIRM(dml_data_iivm, Lasso(), LogisticRegression())
msg = (
'Incompatible data. To fit an IRM model with DML exactly one binary variable with values 0 and 1 '
'needs to be specified as treatment variable.')
df_irm = dml_data_irm.data.copy()
df_irm['d'] = df_irm['d'] * 2
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIRM(DoubleMLData(df_irm, 'y', 'd'), Lasso(),
LogisticRegression())
df_irm = dml_data_irm.data.copy()
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIRM(DoubleMLData(df_irm, 'y', ['d', 'X1']), Lasso(),
LogisticRegression())
msg = (
'Incompatible data. To fit an IIVM model with DML exactly one binary variable with values 0 and 1 '
'needs to be specified as treatment variable.')
df_iivm = dml_data_iivm.data.copy()
df_iivm['d'] = df_iivm['d'] * 2
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIIVM(DoubleMLData(df_iivm, 'y', 'd', z_cols='z'), Lasso(),
LogisticRegression(), LogisticRegression())
df_iivm = dml_data_iivm.data.copy()
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIIVM(DoubleMLData(df_iivm, 'y', ['d', 'X1'], z_cols='z'),
Lasso(), LogisticRegression(), LogisticRegression())
msg = (
'Incompatible data. To fit an IIVM model with DML exactly one binary variable with values 0 and 1 '
'needs to be specified as instrumental variable.')
df_iivm = dml_data_iivm.data.copy()
df_iivm['z'] = df_iivm['z'] * 2
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIIVM(DoubleMLData(df_iivm, 'y', 'd', z_cols='z'), Lasso(),
LogisticRegression(), LogisticRegression())
df_iivm = dml_data_iivm.data.copy()
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIIVM(DoubleMLData(df_iivm, 'y', 'd', z_cols=['z', 'X1']),
Lasso(), LogisticRegression(), LogisticRegression())
def test_doubleml_exception_subgroups():
msg = 'Invalid subgroups True. subgroups must be of type dictionary.'
with pytest.raises(TypeError, match=msg):
_ = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(), LogisticRegression(),
subgroups=True)
msg = "Invalid subgroups {'abs': True}. subgroups must be a dictionary with keys always_takers and never_takers."
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(), LogisticRegression(),
subgroups={'abs': True})
msg = ("Invalid subgroups {'always_takers': True, 'never_takers': False, 'abs': 5}. "
"subgroups must be a dictionary with keys always_takers and never_takers.")
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(), LogisticRegression(),
subgroups={'always_takers': True, 'never_takers': False, 'abs': 5})
msg = ("Invalid subgroups {'always_takers': True}. "
"subgroups must be a dictionary with keys always_takers and never_takers.")
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(), LogisticRegression(),
subgroups={'always_takers': True})
msg = r"subgroups\['always_takers'\] must be True or False. Got 5."
with pytest.raises(TypeError, match=msg):
_ = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(), LogisticRegression(),
subgroups={'always_takers': 5, 'never_takers': False})
msg = r"subgroups\['never_takers'\] must be True or False. Got 5."
with pytest.raises(TypeError, match=msg):
_ = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(), LogisticRegression(),
subgroups={'always_takers': True, 'never_takers': 5})
def test_doubleml_exception_trimming_rule():
msg = 'Invalid trimming_rule discard. Valid trimming_rule truncate.'
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIRM(dml_data_irm,
Lasso(),
LogisticRegression(),
trimming_rule='discard')
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIIVM(dml_data_iivm,
Lasso(),
LogisticRegression(),
LogisticRegression(),
trimming_rule='discard')
def test_doubleml_exception_learner():
err_msg_prefix = 'Invalid learner provided for ml_g: '
warn_msg_prefix = 'Learner provided for ml_g is probably invalid: '
msg = err_msg_prefix + 'provide an instance of a learner instead of a class.'
with pytest.raises(TypeError, match=msg):
_ = DoubleMLPLR(dml_data, Lasso, ml_m)
msg = err_msg_prefix + r'BaseEstimator\(\) has no method .fit\(\).'
with pytest.raises(TypeError, match=msg):
_ = DoubleMLPLR(dml_data, BaseEstimator(), ml_m)
# msg = err_msg_prefix + r'_DummyNoSetParams\(\) has no method .set_params\(\).'
with pytest.raises(TypeError):
_ = DoubleMLPLR(dml_data, _DummyNoSetParams(), ml_m)
# msg = err_msg_prefix + r'_DummyNoSetParams\(\) has no method .get_params\(\).'
with pytest.raises(TypeError):
_ = DoubleMLPLR(dml_data, _DummyNoGetParams(), ml_m)
# msg = 'Learner provided for ml_m is probably invalid: ' + r'_DummyNoClassifier\(\) is \(probably\) no classifier.'
with pytest.warns(UserWarning):
_ = DoubleMLIRM(dml_data_irm, Lasso(), _DummyNoClassifier())
# ToDo: Currently for ml_g (and others) we only check whether the learner can be identified as regressor. However,
# we do not check whether it can instead be identified as classifier, which could be used to throw an error.
msg = warn_msg_prefix + r'LogisticRegression\(\) is \(probably\) no regressor.'
with pytest.warns(UserWarning, match=msg):
_ = DoubleMLPLR(dml_data, LogisticRegression(), Lasso())
# we allow classifiers for ml_m in PLR, but only for binary treatment variables
msg = (r'The ml_m learner LogisticRegression\(\) was identified as classifier '
'but at least one treatment variable is not binary with values 0 and 1.')
with pytest.raises(ValueError, match=msg):
_ = DoubleMLPLR(dml_data, Lasso(), LogisticRegression())
# we allow classifiers for ml_g for binary treatment variables in IRM
msg = (r'The ml_g learner LogisticRegression\(\) was identified as classifier '
'but the outcome variable is not binary with values 0 and 1.')
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIRM(dml_data_irm, LogisticRegression(), LogisticRegression())
# we allow classifiers for ml_g for binary treatment variables in IRM
msg = (r'The ml_g learner LogisticRegression\(\) was identified as classifier '
'but the outcome variable is not binary with values 0 and 1.')
with pytest.raises(ValueError, match=msg):
_ = DoubleMLIIVM(dml_data_iivm, LogisticRegression(), LogisticRegression(), LogisticRegression())
# construct a classifier which is not identifiable as classifier via is_classifier by sklearn
# it then predicts labels and therefore an exception will be thrown
log_reg = LogisticRegression()
log_reg._estimator_type = None
msg = (r'Learner provided for ml_m is probably invalid: LogisticRegression\(\) is \(probably\) neither a regressor '
'nor a classifier. Method predict is used for prediction.')
with pytest.warns(UserWarning, match=msg):
dml_plr_hidden_classifier = DoubleMLPLR(dml_data_irm, Lasso(), log_reg)
msg = (r'For the binary treatment variable d, predictions obtained with the ml_m learner LogisticRegression\(\) '
'are also observed to be binary with values 0 and 1. Make sure that for classifiers probabilities and not '
'labels are predicted.')
with pytest.raises(ValueError, match=msg):
dml_plr_hidden_classifier.fit()
# construct a classifier which is not identifiable as classifier via is_classifier by sklearn
# it then predicts labels and therefore an exception will be thrown
# whether predict() or predict_proba() is being called can also be manipulated via the unrelated max_iter variable
log_reg = LogisticRegressionManipulatedPredict()
log_reg._estimator_type = None
msg = (r'Learner provided for ml_g is probably invalid: LogisticRegressionManipulatedPredict\(\) is \(probably\) '
'neither a regressor nor a classifier. Method predict is used for prediction.')
with pytest.warns(UserWarning, match=msg):
dml_irm_hidden_classifier = DoubleMLIRM(dml_data_irm_binary_outcome,
log_reg, LogisticRegression())
msg = (r'For the binary outcome variable y, predictions obtained with the ml_g learner '
r'LogisticRegressionManipulatedPredict\(\) are also observed to be binary with values 0 and 1. Make sure '
'that for classifiers probabilities and not labels are predicted.')
with pytest.raises(ValueError, match=msg):
dml_irm_hidden_classifier.fit()
with pytest.raises(ValueError, match=msg):
dml_irm_hidden_classifier.set_ml_nuisance_params('ml_g0', 'd', {'max_iter': 314})
dml_irm_hidden_classifier.fit()
msg = (r'Learner provided for ml_g is probably invalid: LogisticRegressionManipulatedPredict\(\) is \(probably\) '
'neither a regressor nor a classifier. Method predict is used for prediction.')
with pytest.warns(UserWarning, match=msg):
dml_iivm_hidden_classifier = DoubleMLIIVM(dml_data_iivm_binary_outcome,
log_reg, LogisticRegression(), LogisticRegression())
msg = (r'For the binary outcome variable y, predictions obtained with the ml_g learner '
r'LogisticRegressionManipulatedPredict\(\) are also observed to be binary with values 0 and 1. Make sure '
'that for classifiers probabilities and not labels are predicted.')
with pytest.raises(ValueError, match=msg):
dml_iivm_hidden_classifier.fit()
with pytest.raises(ValueError, match=msg):
dml_iivm_hidden_classifier.set_ml_nuisance_params('ml_g0', 'd', {'max_iter': 314})
dml_iivm_hidden_classifier.fit()
from doubleml import DoubleMLPLR, DoubleMLIRM, DoubleMLIIVM, DoubleMLPLIV
from doubleml.datasets import make_plr_CCDDHNR2018, make_irm_data, make_pliv_CHS2015, make_iivm_data
from sklearn.linear_model import Lasso, LogisticRegression
np.random.seed(3141)
dml_data_plr = make_plr_CCDDHNR2018(n_obs=100)
dml_data_pliv = make_pliv_CHS2015(n_obs=100, dim_z=1)
dml_data_irm = make_irm_data(n_obs=100)
dml_data_iivm = make_iivm_data(n_obs=100)
dml_plr = DoubleMLPLR(dml_data_plr, Lasso(), Lasso())
dml_pliv = DoubleMLPLIV(dml_data_pliv, Lasso(), Lasso(), Lasso())
dml_irm = DoubleMLIRM(dml_data_irm, Lasso(), LogisticRegression())
dml_iivm = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(),
LogisticRegression())
@pytest.mark.ci
@pytest.mark.parametrize('dml_obj, cls', [(dml_plr, DoubleMLPLR),
(dml_pliv, DoubleMLPLIV),
(dml_irm, DoubleMLIRM),
(dml_iivm, DoubleMLIIVM)])
def test_plr_return_types(dml_obj, cls):
# ToDo: A second test case with multiple treatment variables would be helpful
assert isinstance(dml_obj.__str__(), str)
assert isinstance(dml_obj.summary, pd.DataFrame)
assert isinstance(dml_obj.draw_sample_splitting(), cls)
assert isinstance(dml_obj.set_sample_splitting(dml_obj.smpls), cls)
assert isinstance(dml_obj.fit(), cls)
assert isinstance(dml_obj.__str__(),
from doubleml import DoubleMLPLR, DoubleMLIRM, DoubleMLIIVM, DoubleMLPLIV
from doubleml.datasets import make_plr_CCDDHNR2018, make_irm_data, make_pliv_CHS2015, make_iivm_data
from sklearn.linear_model import Lasso, LogisticRegression
np.random.seed(3141)
dml_data_plr = make_plr_CCDDHNR2018(n_obs=100)
dml_data_pliv = make_pliv_CHS2015(n_obs=100, dim_z=1)
dml_data_irm = make_irm_data(n_obs=100)
dml_data_iivm = make_iivm_data(n_obs=100)
dml_plr = DoubleMLPLR(dml_data_plr, Lasso(), Lasso())
dml_pliv = DoubleMLPLIV(dml_data_pliv, Lasso(), Lasso(), Lasso())
dml_irm = DoubleMLIRM(dml_data_irm, Lasso(), LogisticRegression())
dml_iivm = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(),
LogisticRegression())
def _assert_resampling_default_settings(dml_obj):
assert dml_obj.n_folds == 5
assert dml_obj.n_rep == 1
assert dml_obj.draw_sample_splitting
assert dml_obj.apply_cross_fitting
@pytest.mark.ci
def test_plr_defaults():
_assert_resampling_default_settings(dml_plr)
assert dml_plr.score == 'partialling out'
assert dml_plr.dml_procedure == 'dml2'
from sklearn.linear_model import Lasso, LogisticRegression
np.random.seed(3141)
dml_data_plr = make_plr_CCDDHNR2018(n_obs=100)
dml_data_pliv = make_pliv_CHS2015(n_obs=100, dim_z=1)
dml_data_irm = make_irm_data(n_obs=100)
dml_data_iivm = make_iivm_data(n_obs=100)
dml_plr = DoubleMLPLR(dml_data_plr, Lasso(), Lasso())
dml_plr.fit()
dml_pliv = DoubleMLPLIV(dml_data_pliv, Lasso(), Lasso(), Lasso())
dml_pliv.fit()
dml_irm = DoubleMLIRM(dml_data_irm, Lasso(), LogisticRegression())
dml_irm.fit()
dml_iivm = DoubleMLIIVM(dml_data_iivm, Lasso(), LogisticRegression(),
LogisticRegression())
dml_iivm.fit()
# fit models with callable scores
plr_score = dml_plr._score_elements
dml_plr_callable_score = DoubleMLPLR(dml_data_plr,
Lasso(),
Lasso(),
score=plr_score,
draw_sample_splitting=False)
dml_plr_callable_score.set_sample_splitting(dml_plr.smpls)
dml_plr_callable_score.fit(store_predictions=True)
irm_score = dml_irm._score_elements
dml_irm_callable_score = DoubleMLIRM(dml_data_irm,
Lasso(),