causal_falsify.algorithms.mint module

class causal_falsify.algorithms.mint.MINT(feature_representation='linear', feature_representation_params={}, binary_treatment=False, binary_outcome=False, min_samples_per_env=25, independence_test_args={}, n_bootstraps=1000)

Bases: AbstractFalsificationAlgorithm

get_diagnostics()

Returns quality of fit for nuisance models per environment from the most recent test() call.

Returns:

Diagnostics for model fit, including source labels and mean squared errors for outcome and treatment models.

Return type:

dict

get_feature_representation()

Returns function that outputs feature representation

Raises:

ValueError: If invalid feature_representation or feature_representation_params are provided

Returns:

: Callable function

Feature representation

run_bootstrapped_independence_test(data_x, data_y, resampled_data_x, resampled_data_y)

Runs bootstrapped independence test between data_x and data_y

Returns:

: float

p-value from test

run_independence_test(data_x, data_y)

Runs independence test between data_x and data_y

Returns:

: float

p-value from test

test(data, covariate_vars, treatment_var, outcome_var, source_var)

Perform falsification test for joint test of unconfoundedness and independence of causal mechanisms.

Parameters:

• data (pandas.DataFrame) – DataFrame containing all data from all environments.

• covariate_vars (list of str) – List of covariate column names.

• treatment_var (str) – Name of the treatment column.

• outcome_var (str) – Name of the outcome column.

• source_var (str) – Name of the source/environment column.

Returns:

p-value from the independence test; low p-value implies unmeasured confounding may be present.

Return type:

float