containers.SCSPILLEffects

utils.scspill_helpers.containers.SCSPILLEffects(
    att,
    att_ci,
    cf_mean,
    cf_lower,
    cf_upper,
    spill_mean,
    spill_lower,
    spill_upper,
    ci_level,
    interval_kind='bayesian',
    att_scm=None,
    att_draws=None,
    att_plugin=None,
    propagate_alpha=None,
    n_draws_used=None,
)

Treatment and spillover effects: the contract every model fills.

All paths cover the full sample (pre and post periods). Pre-period rows of the spillover panel are fit residuals by construction, not causal spillovers; the counterfactual band may be NaN where a model defines no interval (cd before treatment).

Attributes

Name	Type	Description
att	float	Average treatment effect on the (headline) treated unit over the post-treatment periods.
att_ci	tuple of float or None	Interval for the ATT at level ci_level; None when the model provides none (cd reports pointwise intervals only).
cf_mean, cf_lower, cf_upper	np.ndarray	Point estimate and pointwise band of the treated unit’s untreated counterfactual path, each shape (T,).
spill_mean, spill_lower, spill_upper	pd.DataFrame	Point estimate and pointwise band of the spillover effects, each (T, N) indexed by time labels with control-label columns.
ci_level	float	Level of all intervals.
interval_kind	{'bayesian', 'cd_inversion', 'cd_published'}	How the intervals were built.
att_scm	float or None	ATT of the no-spillover comparator: the plain synthetic control (sar: the rho = 0 Bayesian horseshoe SCM at alpha_hat; cd: the treated unit’s own SCM equation).
The remaining attributes are set by sar only (None otherwise)
att_draws	np.ndarray	One ATT draw per retained posterior draw, shape (M,).
att_plugin	float	ATT evaluated at the posterior means (alpha_hat, rho_hat).
propagate_alpha	bool	True when effect draws paired (alpha^(m), rho^(m)); False when alpha was held at alpha_hat (the R package’s convention).
n_draws_used	int	Number of posterior draws swept in the effects computation.