PriorSensitivity

class causalpy.checks.prior_sensitivity.PriorSensitivity

Re-fit the experiment with alternative models/priors and compare.

Each alternative is specified as a dict with "name" and "model" keys. The check re-instantiates the experiment for each alternative model and compares the resulting effect summaries.

Parameters:

alternatives (list[dict[str, Any]]) – Each dict must have "name" (str) and "model" (PyMCModel or RegressorMixin) keys.

Examples

>>> import causalpy as cp
>>> check = cp.checks.PriorSensitivity(
...     alternatives=[
...         {"name": "diffuse", "model": cp.pymc_models.LinearRegression(...)},
...         {"name": "tight", "model": cp.pymc_models.LinearRegression(...)},
...     ]
... )

Methods

PriorSensitivity.__init__(alternatives)
PriorSensitivity.run(experiment, context)	Re-fit with each alternative model and compare effect estimates.
PriorSensitivity.validate(experiment)	Verify the experiment uses a Bayesian (PyMC) model.

Attributes

applicable_methods

__init__(alternatives)

Parameters:

alternatives (list[dict[str, Any]])

Return type:

None

classmethod __new__(*args, **kwargs)