mcf_main.ModifiedCausalForest.train

ModifiedCausalForest.train(data_df, exit_after_commonsupport=False)

Build the modified causal forest on the training data.

Parameters:

• data_df (DataFrame) – Data used to compute the causal forest. It must contain information about outcomes, treatment, and features.

• exit_after_commonsupport (Boolean, optional) – If True, return the training-result dictionary after common-support processing. Data preparation and sample splitting still run, but local centering and causal forest training are skipped. To apply the fitted common-support rules to new data, call predict() with ‘exit_after_commonsupport=True’. If ‘cs_type’ is 0, common-support processing is skipped and the method still returns before forest training. The default is False.

Returns:

results – Training samples and common-support information. The same keys are returned for both values of ‘exit_after_commonsupport’.

’tree_df’DataFrame

Processed sample for building the trees.

’fill_y_df’DataFrame

Processed sample for filling the trees with outcome values.

’common_support_probabilities_tree’DataFrame or None

Common-support table for the sample used to build the trees, before trimming. Contains treatment probabilities, the observation identifier, and ‘on_support’ (1 if retained, 0 if excluded). None if ‘cs_type’ is 0 or ‘_int_with_output’ is False.

’common_support_probabilities_fill_y’DataFrame or None

Same information for the sample used to fill the trees, before trimming. None if ‘cs_type’ is 0 or ‘_int_with_output’ is False.

’path_output’Path or None

Directory in which output is saved. None if _int_with_output is False.

Return type:

Dictionary.