def test_SensitivityRandomCause():
y, X, treatment, tau, b, e = synthetic_data(mode=1,
n=100000,
p=NUM_FEATURES,
sigma=1.0)
# generate the dataset format for sensitivity analysis
INFERENCE_FEATURES = ["feature_" + str(i) for i in range(NUM_FEATURES)]
df = pd.DataFrame(X, columns=INFERENCE_FEATURES)
df[TREATMENT_COL] = treatment
df[OUTCOME_COL] = y
df[SCORE_COL] = e
# calling the Base XLearner class and return the sensitivity analysis summary report
learner = BaseXLearner(LinearRegression())
sens = SensitivityRandomCause(
df=df,
inference_features=INFERENCE_FEATURES,
p_col=SCORE_COL,
treatment_col=TREATMENT_COL,
outcome_col=OUTCOME_COL,
learner=learner,
)
sens_summary = sens.summary(method="Random Cause")
print(sens_summary)
def test_SensitivitySelectionBias():
y, X, treatment, tau, b, e = synthetic_data(mode=1,
n=100000,
p=NUM_FEATURES,
sigma=1.0)
# generate the dataset format for sensitivity analysis
INFERENCE_FEATURES = ["feature_" + str(i) for i in range(NUM_FEATURES)]
df = pd.DataFrame(X, columns=INFERENCE_FEATURES)
df[TREATMENT_COL] = treatment
df[OUTCOME_COL] = y
df[SCORE_COL] = e
# calling the Base XLearner class and return the sensitivity analysis summary report
learner = BaseXLearner(LinearRegression())
sens = SensitivitySelectionBias(
df,
INFERENCE_FEATURES,
p_col=SCORE_COL,
treatment_col=TREATMENT_COL,
outcome_col=OUTCOME_COL,
learner=learner,
confound="alignment",
alpha_range=None,
)
lls_bias_alignment, partial_rsqs_bias_alignment = sens.causalsens()
print(lls_bias_alignment, partial_rsqs_bias_alignment)
# Plot the results by confounding vector and plot Confidence Intervals for ATE
sens.plot(lls_bias_alignment, ci=True)
def test_Sensitivity():
y, X, treatment, tau, b, e = synthetic_data(mode=1,
n=100000,
p=NUM_FEATURES,
sigma=1.0)
# generate the dataset format for sensitivity analysis
INFERENCE_FEATURES = ['feature_' + str(i) for i in range(NUM_FEATURES)]
df = pd.DataFrame(X, columns=INFERENCE_FEATURES)
df[TREATMENT_COL] = treatment
df[OUTCOME_COL] = y
df[SCORE_COL] = e
# calling the Base XLearner class and return the sensitivity analysis summary report
learner = BaseXLearner(LinearRegression())
sens = Sensitivity(df=df,
inference_features=INFERENCE_FEATURES,
p_col=SCORE_COL,
treatment_col=TREATMENT_COL,
outcome_col=OUTCOME_COL,
learner=learner)
# check the sensitivity summary report
sens_summary = sens.sensitivity_analysis(methods=[
'Placebo Treatment', 'Random Cause', 'Subset Data', 'Random Replace',
'Selection Bias'
],
sample_size=0.5)
print(sens_summary)
def test_alignment_att():
y, X, treatment, tau, b, e = synthetic_data(mode=1,
n=100000,
p=NUM_FEATURES,
sigma=1.0)
alpha = np.quantile(y, 0.25)
adj = alignment_att(alpha, e, treatment)
assert y.shape == adj.shape
def test_synthetic_data():
y, X, treatment, tau, b, e = synthetic_data(mode=1,
n=N_SAMPLE,
p=8,
sigma=.1)
assert (y.shape[0] == X.shape[0] and y.shape[0] == treatment.shape[0]
and y.shape[0] == tau.shape[0] and y.shape[0] == b.shape[0]
and y.shape[0] == e.shape[0])
y, X, treatment, tau, b, e = synthetic_data(mode=2,
n=N_SAMPLE,
p=8,
sigma=.1)
assert (y.shape[0] == X.shape[0] and y.shape[0] == treatment.shape[0]
and y.shape[0] == tau.shape[0] and y.shape[0] == b.shape[0]
and y.shape[0] == e.shape[0])
y, X, treatment, tau, b, e = synthetic_data(mode=3,
n=N_SAMPLE,
p=8,
sigma=.1)
assert (y.shape[0] == X.shape[0] and y.shape[0] == treatment.shape[0]
and y.shape[0] == tau.shape[0] and y.shape[0] == b.shape[0]
and y.shape[0] == e.shape[0])
y, X, treatment, tau, b, e = synthetic_data(mode=4,
n=N_SAMPLE,
p=8,
sigma=.1)
assert (y.shape[0] == X.shape[0] and y.shape[0] == treatment.shape[0]
and y.shape[0] == tau.shape[0] and y.shape[0] == b.shape[0]
and y.shape[0] == e.shape[0])
def _generate_data():
if not generated:
np.random.seed(RANDOM_SEED)
data = synthetic_data(mode=1, n=N_SAMPLE, p=8, sigma=.1)
return data
'''
Day 52
Uber CausalML
'''
from causalml.inference import LRSRegressor
from causalml.inference import XGBTRegressor, MLPTRegressor
from causalml.inference import BaseXRegressor
from causalml.dataset import synthetic_data
y, X, treatment, _ = synthetic_data(mode=1, n=1000, p=5, sigma=1.0)
lr = LRSRegressor()
te, lb, ub = lr.estimate_ate(X, treatment, y)
logger.info(
'Average Treatment Effect (Linear Regression): {:.2f} ({:.2f}, {:.2f})'.
format(te, lb, ub))
xg = XGBTRegressor(random_state=42)
te, lb, ub = xg.estimate_ate(X, treatment, y)
logger.info(
'Average Treatment Effect (XGBoost): {:.2f} ({:.2f}, {:.2f})'.format(
te, lb, ub))
nn = MLPTRegressor(hidden_layer_sizes=(10, 10),
learning_rate_init=.1,
early_stopping=True,
random_state=42)
te, lb, ub = nn.estimate_ate(X, treatment, y)
logger.info(
