def dml_iivm_pyvsr_fixture(generate_data_iivm, idx, score, dml_procedure):
boot_methods = ['normal']
n_folds = 2
# collect data
data = generate_data_iivm[idx]
X_cols = data.columns[data.columns.str.startswith('X')].tolist()
# Set machine learning methods for m & gg
learner_classif = LogisticRegression(penalty='none', solver='newton-cg')
learner_reg = LinearRegression()
ml_g = clone(learner_reg)
ml_m = clone(learner_classif)
ml_r = clone(learner_classif)
obj_dml_data = dml.DoubleMLData(data, 'y', ['d'], X_cols, 'z')
dml_iivm_obj = dml.DoubleMLIIVM(obj_dml_data,
ml_g,
ml_m,
ml_r,
n_folds,
dml_procedure=dml_procedure)
np.random.seed(3141)
dml_iivm_obj.fit()
# fit the DML model in R
all_train, all_test = export_smpl_split_to_r(dml_iivm_obj.smpls[0])
r_dataframe = pandas2ri.py2rpy(data)
res_r = r_IIVM(r_dataframe, score, dml_procedure, all_train, all_test)
res_dict = {
'coef_py': dml_iivm_obj.coef,
'coef_r': res_r[0],
'se_py': dml_iivm_obj.se,
'se_r': res_r[1]
}
return res_dict
def dml_iivm_fixture(generate_data_iivm, learner_g, learner_m, learner_r,
score, dml_procedure, tune_on_folds):
par_grid = {
'ml_g': get_par_grid(learner_g),
'ml_m': get_par_grid(learner_m),
'ml_r': get_par_grid(learner_r)
}
n_folds_tune = 4
boot_methods = ['normal']
n_folds = 2
n_rep_boot = 491
# collect data
data = generate_data_iivm
x_cols = data.columns[data.columns.str.startswith('X')].tolist()
# Set machine learning methods for m, g & r
ml_g = clone(learner_g)
ml_m = clone(learner_m)
ml_r = clone(learner_r)
np.random.seed(3141)
obj_dml_data = dml.DoubleMLData(data, 'y', ['d'], x_cols, 'z')
dml_iivm_obj = dml.DoubleMLIIVM(obj_dml_data,
ml_g,
ml_m,
ml_r,
n_folds,
dml_procedure=dml_procedure)
# tune hyperparameters
_ = dml_iivm_obj.tune(par_grid,
tune_on_folds=tune_on_folds,
n_folds_tune=n_folds_tune)
dml_iivm_obj.fit()
np.random.seed(3141)
y = data['y'].values
x = data.loc[:, x_cols].values
d = data['d'].values
z = data['z'].values
resampling = KFold(n_splits=n_folds, shuffle=True)
smpls = [(train, test) for train, test in resampling.split(x)]
if tune_on_folds:
g0_params, g1_params, m_params, r0_params, r1_params = \
tune_nuisance_iivm(y, x, d, z,
clone(learner_m), clone(learner_g), clone(learner_r), smpls,
n_folds_tune,
par_grid['ml_g'], par_grid['ml_m'], par_grid['ml_r'])
g_hat0, g_hat1, m_hat, r_hat0, r_hat1 = \
fit_nuisance_iivm(y, x, d, z,
clone(learner_m), clone(learner_g), clone(learner_r), smpls,
g0_params, g1_params, m_params, r0_params, r1_params)
else:
xx = [(np.arange(data.shape[0]), np.array([]))]
g0_params, g1_params, m_params, r0_params, r1_params = \
tune_nuisance_iivm(y, x, d, z,
clone(learner_m), clone(learner_g), clone(learner_r), xx,
n_folds_tune,
par_grid['ml_g'], par_grid['ml_m'], par_grid['ml_r'])
g_hat0, g_hat1, m_hat, r_hat0, r_hat1 = \
fit_nuisance_iivm(y, x, d, z,
clone(learner_m), clone(learner_g), clone(learner_r), smpls,
g0_params * n_folds, g1_params * n_folds, m_params * n_folds,
r0_params * n_folds, r1_params * n_folds)
if dml_procedure == 'dml1':
res_manual, se_manual = iivm_dml1(y, x, d, z, g_hat0, g_hat1, m_hat,
r_hat0, r_hat1, smpls, score)
else:
assert dml_procedure == 'dml2'
res_manual, se_manual = iivm_dml2(y, x, d, z, g_hat0, g_hat1, m_hat,
r_hat0, r_hat1, smpls, score)
res_dict = {
'coef': dml_iivm_obj.coef,
'coef_manual': res_manual,
'se': dml_iivm_obj.se,
'se_manual': se_manual,
'boot_methods': boot_methods
}
for bootstrap in boot_methods:
np.random.seed(3141)
boot_theta, boot_t_stat = boot_iivm(res_manual, y, d, z, g_hat0,
g_hat1, m_hat, r_hat0, r_hat1,
smpls, score, se_manual, bootstrap,
n_rep_boot, dml_procedure)
np.random.seed(3141)
dml_iivm_obj.bootstrap(method=bootstrap, n_rep_boot=n_rep_boot)
res_dict['boot_coef' + bootstrap] = dml_iivm_obj.boot_coef
res_dict['boot_t_stat' + bootstrap] = dml_iivm_obj.boot_t_stat
res_dict['boot_coef' + bootstrap + '_manual'] = boot_theta
res_dict['boot_t_stat' + bootstrap + '_manual'] = boot_t_stat
return res_dict
def dml_iivm_fixture(generate_data_iivm, learner, score, dml_procedure,
trimming_threshold):
boot_methods = ['normal']
n_folds = 2
n_rep_boot = 491
# collect data
data = generate_data_iivm
x_cols = data.columns[data.columns.str.startswith('X')].tolist()
# Set machine learning methods for m & g
ml_g = clone(learner[0])
ml_m = clone(learner[1])
ml_r = clone(learner[1])
np.random.seed(3141)
obj_dml_data = dml.DoubleMLData(data, 'y', ['d'], x_cols, 'z')
dml_iivm_obj = dml.DoubleMLIIVM(obj_dml_data,
ml_g,
ml_m,
ml_r,
n_folds,
dml_procedure=dml_procedure,
trimming_threshold=trimming_threshold)
dml_iivm_obj.fit()
np.random.seed(3141)
y = data['y'].values
x = data.loc[:, x_cols].values
d = data['d'].values
z = data['z'].values
n_obs = len(y)
all_smpls = draw_smpls(n_obs, n_folds)
res_manual = fit_iivm(y,
x,
d,
z,
clone(learner[0]),
clone(learner[1]),
clone(learner[1]),
all_smpls,
dml_procedure,
score,
trimming_threshold=trimming_threshold)
res_dict = {
'coef': dml_iivm_obj.coef,
'coef_manual': res_manual['theta'],
'se': dml_iivm_obj.se,
'se_manual': res_manual['se'],
'boot_methods': boot_methods
}
for bootstrap in boot_methods:
np.random.seed(3141)
boot_theta, boot_t_stat = boot_iivm(
y, d, z, res_manual['thetas'], res_manual['ses'],
res_manual['all_g_hat0'], res_manual['all_g_hat1'],
res_manual['all_m_hat'], res_manual['all_r_hat0'],
res_manual['all_r_hat1'], all_smpls, score, bootstrap, n_rep_boot)
np.random.seed(3141)
dml_iivm_obj.bootstrap(method=bootstrap, n_rep_boot=n_rep_boot)
res_dict['boot_coef' + bootstrap] = dml_iivm_obj.boot_coef
res_dict['boot_t_stat' + bootstrap] = dml_iivm_obj.boot_t_stat
res_dict['boot_coef' + bootstrap + '_manual'] = boot_theta
res_dict['boot_t_stat' + bootstrap + '_manual'] = boot_t_stat
return res_dict
def dml_iivm_fixture(generate_data_iivm, learner, score, dml_procedure,
trimming_threshold):
boot_methods = ['normal']
n_folds = 2
n_rep_boot = 491
# collect data
data = generate_data_iivm
x_cols = data.columns[data.columns.str.startswith('X')].tolist()
# Set machine learning methods for m & g
ml_g = clone(learner[1])
ml_m = clone(learner[0])
ml_r = clone(learner[0])
np.random.seed(3141)
obj_dml_data = dml.DoubleMLData(data, 'y', ['d'], x_cols, 'z')
dml_iivm_obj = dml.DoubleMLIIVM(obj_dml_data,
ml_g,
ml_m,
ml_r,
n_folds,
dml_procedure=dml_procedure,
trimming_threshold=trimming_threshold)
dml_iivm_obj.fit()
np.random.seed(3141)
y = data['y'].values
x = data.loc[:, x_cols].values
d = data['d'].values
z = data['z'].values
resampling = KFold(n_splits=n_folds, shuffle=True)
smpls = [(train, test) for train, test in resampling.split(x)]
g_hat0, g_hat1, m_hat, r_hat0, r_hat1 = fit_nuisance_iivm(
y,
x,
d,
z,
clone(learner[0]),
clone(learner[1]),
clone(learner[0]),
smpls,
trimming_threshold=trimming_threshold)
if dml_procedure == 'dml1':
res_manual, se_manual = iivm_dml1(y, x, d, z, g_hat0, g_hat1, m_hat,
r_hat0, r_hat1, smpls, score)
else:
assert dml_procedure == 'dml2'
res_manual, se_manual = iivm_dml2(y, x, d, z, g_hat0, g_hat1, m_hat,
r_hat0, r_hat1, smpls, score)
res_dict = {
'coef': dml_iivm_obj.coef,
'coef_manual': res_manual,
'se': dml_iivm_obj.se,
'se_manual': se_manual,
'boot_methods': boot_methods
}
for bootstrap in boot_methods:
np.random.seed(3141)
boot_theta, boot_t_stat = boot_iivm(res_manual, y, d, z, g_hat0,
g_hat1, m_hat, r_hat0, r_hat1,
smpls, score, se_manual, bootstrap,
n_rep_boot, dml_procedure)
np.random.seed(3141)
dml_iivm_obj.bootstrap(method=bootstrap, n_rep_boot=n_rep_boot)
res_dict['boot_coef' + bootstrap] = dml_iivm_obj.boot_coef
res_dict['boot_t_stat' + bootstrap] = dml_iivm_obj.boot_t_stat
res_dict['boot_coef' + bootstrap + '_manual'] = boot_theta
res_dict['boot_t_stat' + bootstrap + '_manual'] = boot_t_stat
return res_dict
def dml_iivm_classifier_fixture(generate_data_iivm_binary, learner, score,
dml_procedure, trimming_threshold):
boot_methods = ['normal']
n_folds = 2
n_rep_boot = 491
# collect data
(x, y, d, z) = generate_data_iivm_binary
# Set machine learning methods for m & g
ml_g = clone(learner[0])
ml_m = clone(learner[1])
ml_r = clone(learner[1])
np.random.seed(3141)
obj_dml_data = dml.DoubleMLData.from_arrays(x, y, d, z)
dml_iivm_obj = dml.DoubleMLIIVM(obj_dml_data,
ml_g,
ml_m,
ml_r,
n_folds,
dml_procedure=dml_procedure,
trimming_threshold=trimming_threshold)
dml_iivm_obj.fit()
np.random.seed(3141)
n_obs = len(y)
all_smpls = draw_smpls(n_obs, n_folds)
res_manual = fit_iivm(y,
x,
d,
z,
clone(learner[0]),
clone(learner[1]),
clone(learner[1]),
all_smpls,
dml_procedure,
score,
trimming_threshold=trimming_threshold)
res_dict = {
'coef': dml_iivm_obj.coef,
'coef_manual': res_manual['theta'],
'se': dml_iivm_obj.se,
'se_manual': res_manual['se'],
'boot_methods': boot_methods
}
for bootstrap in boot_methods:
np.random.seed(3141)
boot_theta, boot_t_stat = boot_iivm(
y, d, z, res_manual['thetas'], res_manual['ses'],
res_manual['all_g_hat0'], res_manual['all_g_hat1'],
res_manual['all_m_hat'], res_manual['all_r_hat0'],
res_manual['all_r_hat1'], all_smpls, score, bootstrap, n_rep_boot)
np.random.seed(3141)
dml_iivm_obj.bootstrap(method=bootstrap, n_rep_boot=n_rep_boot)
res_dict['boot_coef' + bootstrap] = dml_iivm_obj.boot_coef
res_dict['boot_t_stat' + bootstrap] = dml_iivm_obj.boot_t_stat
res_dict['boot_coef' + bootstrap + '_manual'] = boot_theta
res_dict['boot_t_stat' + bootstrap + '_manual'] = boot_t_stat
return res_dict