causal_falsify.algorithms.abstract module

class causal_falsify.algorithms.abstract.AbstractFalsificationAlgorithm[source]

Bases: ABC

abstract test(data, covariate_vars, treatment_var, outcome_var, source_var)[source]

Run the falsification test to test the null hypothesis of no unmeasured confounding in multi-source observational data.

Parameters:

  • data (pd.DataFrame) – Input dataset.

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

  • treatment_var (str) – Treatment variable column name.

  • outcome_var (str) – Outcome variable column name.

  • source_var (str) – Source/environment indicator column.

Returns:

p-value from the test.

Return type:

float