mcf_functions.ModifiedCausalForest.blinder_iates#

ModifiedCausalForest.blinder_iates(data_df, blind_var_x_protected_name=None, blind_var_x_policy_name=None, blind_var_x_unrestricted_name=None, blind_weights_of_blind=None, blind_obs_ref_data=50, blind_seed=123456)#

Compute IATEs that causally depend less on protected variables.

WARNING This method is deprecated and will be removed in future versions. Use the method fairscores of the OptimalPolicy class instead.

Parameters

  • data_df (DataFrame.) – Contains data needed to predict() the various adjusted IATES.

  • blind_var_x_protected_name (List of strings (or None), optional) – Names of protected variables. Names that are explicitly denoted as blind_var_x_unrestricted_name or as blind_var_x_policy_name and used to compute IATEs will be automatically added to this list. Default is None.

  • blind_var_x_policy_name (List of strings (or None), optional) – Names of decision variables. Default is None.

  • blind_var_x_unrestricted_name (List of strings (or None), optional) – Names of unrestricted variables. Default is None.

  • blind_weights_of_blind (Tuple of float (or None), optional) – Weights to compute weighted means of blinded and unblinded IATEs. Between 0 and 1. 1 implies all weight goes to fully blinded IATEs.

  • blind_obs_ref_data (Integer (or None), optional) – Number of observations to be used for blinding. Runtime of programme is almost linear in this parameter. Default is 50.

  • blind_seed (Integer, optional.) – Seed for the random selection of the reference data. Default is 123456.

Returns

  • blinded_dic (Dictionary.) – Contains potential outcomes that do not fully depend on some protected attributes (and non-modified IATE).

  • data_on_support_df (DataFrame.) – Features that are on the common support.

  • var_x_policy_ord_name (List of strings) – Ordered variables to be used to build the decision rules.

  • var_x_policy_unord_name (List of strings.) – Unordered variables to be used to build the decision rules.

  • var_x_blind_ord_name (List of strings) – Ordered variables to be used to blind potential outcomes.

  • var_x_blind_unord_name (List of strings.) – Unordered variables to be used to blind potential outcomes.

  • outpath (Pathlib object) – Location of directory in which output is saved.