CounterfactualEvaluator.evaluate_baseline#

CounterfactualEvaluator.evaluate_baseline(target_values)[source]#

Evaluate every node on one set of intervention values, over the full date axis.

Parameters:
target_valuesnp.ndarray

Values for the intervention target, shape (n_dates, *extra_shape). Under the default spend target this is the actual spend; under "scale" it is the mask, and ones reproduce the factual baseline.

Returns:
dict

Per response variable, an array of shape (n_samples, n_dates, *non_date_dims).