causalis.scenarios.synthetic_control.model

Module Contents

Classes

AugmentedSyntheticControl

Ridge-augmented synthetic control with simplex anchor and aggregate-first inference.

Data

ASCM

API

class causalis.scenarios.synthetic_control.model.AugmentedSyntheticControl(*, lambda_aug: float = 1.0, lambda_sc: float = 1e-06, max_iter: int = 2000, tol: float = 1e-09, enforce_sum_to_one_augmented: bool = True, alpha: float = 0.05, conformal_grid_size: int = 401, conformal_grid_min: float | None = None, conformal_grid_max: float | None = None, conformal_grid_scale_mult: float = 6.0, average_att_n_folds: int = 3, compute_average_att_ttest: bool = True, compute_pointwise_conformal: bool = False)

Ridge-augmented synthetic control with simplex anchor and aggregate-first inference.

Notes

Average ATT t-test inference is the default post-treatment inference layer. Pointwise conformal intervals/p-values are optional and can be enabled for dynamic path uncertainty quantification.

Initialization

Initialize ASCM hyperparameters.

Parameters

lambda_aug : float, default=1.0 Ridge regularization for augmented weights. lambda_sc : float, default=1e-6 Numerical regularization for simplex SCM weights. max_iter : int, default=2000 Maximum iterations for constrained optimization routines. tol : float, default=1e-9 Optimization tolerance. enforce_sum_to_one_augmented : bool, default=True Enforce sum-to-one constraint on augmented weights. alpha : float, default=0.05 Default significance level used by estimate() inference. conformal_grid_size : int, default=401 Default number of grid points used in pointwise conformal inversion. conformal_grid_min : float or None, default=None Optional default fixed lower bound for conformal grid. conformal_grid_max : float or None, default=None Optional default fixed upper bound for conformal grid. conformal_grid_scale_mult : float, default=6.0 Default scale multiplier for automatic conformal grid width. average_att_n_folds : int, default=3 Default requested number of folds for average ATT t-test inference. compute_average_att_ttest : bool, default=True Default toggle for average ATT t-test inference in estimate(). compute_pointwise_conformal : bool, default=False Default toggle for pointwise conformal CIs/p-values in estimate().

Raises

ValueError If any hyperparameter is invalid.

fit(data: causalis.data_contracts.panel_data_scm.PanelDataSCM) → causalis.scenarios.synthetic_control.model.AugmentedSyntheticControl

Fit ASCM and compute inference outputs.

Parameters

data : PanelDataSCM Validated synthetic-control panel data.

Returns

AugmentedSyntheticControl Fitted estimator instance.

Raises

ValueError If input type is invalid or panel requirements are violated.

estimate(*, alpha: float | None = None, conformal_grid_size: int | None = None, conformal_grid_min: float | None = None, conformal_grid_max: float | None = None, conformal_grid_scale_mult: float | None = None, average_att_n_folds: int | None = None, compute_average_att_ttest: bool | None = None, compute_pointwise_conformal: bool | None = None) → causalis.data_contracts.panel_estimate.PanelEstimate

Return dynamic-path estimate object.

Parameters

alpha : float or None, default=None Optional per-call significance level override. conformal_grid_size : int or None, default=None Optional per-call pointwise conformal grid size override. conformal_grid_min : float or None, default=None Optional per-call lower conformal grid bound override. conformal_grid_max : float or None, default=None Optional per-call upper conformal grid bound override. conformal_grid_scale_mult : float or None, default=None Optional per-call automatic conformal grid scale override. average_att_n_folds : int or None, default=None Optional per-call fold-count override for average ATT inference. compute_average_att_ttest : bool or None, default=None Optional per-call toggle for average ATT t-test inference. compute_pointwise_conformal : bool or None, default=None Optional per-call toggle for pointwise conformal inference.

Returns

PanelEstimate Dynamic path estimates with pointwise inference fields. Aggregate average ATT t-test outputs are provided in diagnostics and are the default formal inference layer. If pointwise conformal is not computed, pointwise p-values/CIs are returned as NaN placeholders.

Raises

RuntimeError If the model is not fitted.

__repr__() → str

causalis.scenarios.synthetic_control.model.ASCM

None