PlaceboInTime.run#

PlaceboInTime.run(experiment, context=None)[source]#

Run placebo-in-time analysis with hierarchical null model.

Creates n_folds placebo experiments by shifting the treatment time backward. Extracts posterior cumulative impact from each fold, then fits a hierarchical Bayesian model to characterise the status-quo distribution. Compares the actual intervention effect against this null.

When expected_effect_prior was provided at construction, also runs Bayesian assurance simulation.

Can be used standalone (context=None) when experiment_factory was provided, or within a pipeline.

Parameters:

  • experiment (BaseExperiment) – The fitted experiment whose treatment time will be shifted to generate placebo folds.

  • context (PipelineContext | None) – Pipeline context providing experiment_config for re-fits. If None, an explicit experiment_factory must have been supplied at construction time.

Returns:

With passed indicating whether the actual effect is clearly outside the null distribution, and rich metadata including the null samples and optional assurance results.

Return type:

CheckResult