PyMCForecastAdapter

class causalpy.experiments.model_adapter.PyMCForecastAdapter

Adapter for PyMCForecastModel backends.

The wrapped model already speaks CausalPy’s Bayesian conventions (mu/y_hat posterior-predictive output on obs_ind / treated_units coords), so this adapter is pure delegation.

Parameters:

model (PyMCForecastModel) – Wrapped pymc_forecast backend model.

Methods

PyMCForecastAdapter.build(X, y, *[, coords])	Record the fit inputs for sample_posterior().
PyMCForecastAdapter.coefficients(*[, group])	Forecasting models have no design-matrix coefficients.
PyMCForecastAdapter.fit(X, y, *[, coords])	Fit the forecasting model on the pre-period.
PyMCForecastAdapter.predict(X, *[, coords, ...])	Predict in-sample or forecast the counterfactual.
PyMCForecastAdapter.print_coefficients(labels)	Print posterior summaries of the model's scalar parameters.
PyMCForecastAdapter.require_idata()	Return fitted inference-result DataTree or raise an explicit capability error.
PyMCForecastAdapter.sample_posterior(**kwargs)	Fit the forecaster on the recorded inputs.
PyMCForecastAdapter.sample_prior_predictive(...)	Raise: the forecaster exposes no prior-drawing path upstream.
PyMCForecastAdapter.score(X, y, *[, coords])	Score in-sample predictions with the Bayesian \(R^2\).

Attributes

has_posterior	Whether posterior draws are available.
has_prior	Whether prior draws are available on this backend.
idata	Return the model's DataTree when fitted.
is_bayesian	Whether the backend is Bayesian (PyMC or pymc-forecast).
is_built	Whether fit inputs have been recorded via build().
is_ols	Whether the backend is OLS/sklearn.
kind	Backend identifier.
model	The underlying pymc-forecast wrapper.
supports_idata	Whether the backend exposes an inference-result DataTree.
supports_prior_predictive	Whether this backend can sample a prior predictive phase.

__init__(model)

Parameters:

model (PyMCForecastModel)

Return type:

None

classmethod __new__(*args, **kwargs)