pytyche.analysis.evaluate_against_truth¶

pytyche.analysis.evaluate_against_truth(posterior, tree, truth)[source]¶

Evaluate the posterior + tree policy against ground truth.

Sim-mode only: generators pair observed data with a CalibrationTruth; real-data mode has no truth and raises RuntimeError.

Parameters:

posterior (HurdleBCFResult | ContinuousBCFResult | BinaryBCFResult) – One of the three posterior result types, carrying observed data (raises otherwise).
tree (PolicyTreeResult) – The fitted policy whose assignments are evaluated.
truth (CalibrationTruth | None) – Ground truth for the simulated experiment, or None in real-data mode.

Return type:

TruthComparison

Returns:

TruthComparison with all six metrics populated.

Raises:

RuntimeError – When truth is None (real-data mode — nothing to evaluate against).
ValueError – When posterior.observed is None, when the truth lacks the K-appropriate contrast or potential-outcome fields, or when the truth shapes do not match the posterior.