diff_diff.SpilloverDiDResults#

class diff_diff.SpilloverDiDResults[source]#

Bases: DiDResults

Results from a ring-indicator spillover-aware DiD estimation (Butts 2021).

Extends DiDResults with per-ring spillover effect estimates and diagnostic counts for the identifying control population.

att#

Total effect on the treated, tau_total (inherited from DiDResults).

Type:: float

spillover_effects#

Per-ring spillover-on-control estimates. Columns: ["coef", "se", "t_stat", "p_value", "ci_low", "ci_high"]. Index: ring label like "[0, 50)"; for event-study output the index is a MultiIndex over (ring_label, event_time).

Type:: pd.DataFrame, optional

ring_breakpoints#

Sorted distance breakpoints used to construct the rings.

Type:: list of float

d_bar#

Far-away cutoff (Butts Assumption 5). Equals max(ring_breakpoints) unless explicitly overridden.

Type:: float

n_units_ever_in_ring#

Counts of UNIQUE units that appear in each ring AT LEAST ONCE across observed periods. Under staggered timing the ring membership is time-varying, so a unit can be counted in multiple rings (one per period it visits). Under non-staggered timing the rings are unit-static, so this is a clean partition (each unit appears in exactly one ring or the far-away group). Includes treated units in Ring_1 by geometric construction even though the (1 - D_it) factor zeros their stage-2 contribution.

Type:: dict[str, int]

n_far_away_obs#

Number of observations with D_it = 0 AND d_it > d_bar; these observations identify the counterfactual trend (Butts Assumption 5(ii)). Under Wave E.3, on the survey path this count is reported on the EFFECTIVE WEIGHTED ESTIMATION SAMPLE (count_mask = survey_finite_mask = finite_mask AND survey_weights > 0), so zero-weight rows from SurveyDesign.subpopulation() are excluded — matches the Wave E.3 n_obs / n_treated / n_control contract. The upstream _validate_far_away_exists gate still fires on the FULL DOMAIN at fit-time (Assumption 5(ii) requires at least one far-away identifying row); this metadata only reports the count on the active sample.

Type:: int

is_staggered#

True if multiple distinct treatment-onset times were detected.

Type:: bool

event_study#

True if per-event-time ring coefficients were emitted (i.e., self.event_study=True was set on the estimator).

Type:: bool

stage1_n_obs#

Number of observations in the stage-1 untreated-and-unexposed subsample (Omega_0_butts).

Type:: int

Methods

`__init__`(att, se, t_stat, p_value, conf_int, ...)
`print_summary`([alpha])	Print the summary to stdout.
`summary`([alpha])	Extended summary with ATT row, per-event-time direct block, and per-(ring, event-time) spillover block.
`to_dataframe`()	Convert results to a pandas DataFrame.
`to_dict`()	Override to serialize `spillover_effects` DataFrame as list-of-dicts.

Attributes

`alpha`
`anticipation`
`att_dynamic`
`bootstrap_distribution`
`cluster_name`
`coef_var`	SE / abs(ATT).
`coefficients`
`conley_lag_cutoff`
`d_bar`
`event_study`
`event_study_effects`
`fitted_values`
`horizon_max`
`inference_method`
`is_significant`	Check if the ATT is statistically significant at the alpha level.
`is_staggered`
`n_bootstrap`
`n_clusters`
`n_far_away_obs`
`n_psu`
`n_strata`
`n_units_ever_in_ring`
`r_squared`
`reference_period`
`residuals`
`ring_breakpoints`
`significance_stars`	Return significance stars based on p-value.
`spillover_effects`
`stage1_n_obs`
`survey_metadata`
`vcov`
`vcov_type`
`att`
`se`
`t_stat`
`p_value`
`conf_int`
`n_obs`
`n_treated`
`n_control`

__init__(att, se, t_stat, p_value, conf_int, n_obs, n_treated, n_control, alpha=0.05, coefficients=None, vcov=None, residuals=None, fitted_values=None, r_squared=None, inference_method='analytical', n_bootstrap=None, n_clusters=None, bootstrap_distribution=None, survey_metadata=None, vcov_type=None, cluster_name=None, conley_lag_cutoff=None, spillover_effects=None, ring_breakpoints=None, d_bar=None, n_units_ever_in_ring=None, n_far_away_obs=None, is_staggered=None, event_study=None, stage1_n_obs=None, anticipation=None, att_dynamic=None, event_study_effects=None, horizon_max=None, reference_period=None, n_psu=None, n_strata=None)#

Parameters:

att (float)
se (float)
t_stat (float)
p_value (float)
conf_int (Tuple[float, float])
n_obs (int)
n_treated (int)
n_control (int)
alpha (float)
coefficients (Dict[str, float] | None)
vcov (ndarray | None)
residuals (ndarray | None)
fitted_values (ndarray | None)
r_squared (float | None)
inference_method (str)
n_bootstrap (int | None)
n_clusters (int | None)
bootstrap_distribution (ndarray | None)
survey_metadata (Any | None)
vcov_type (str | None)
cluster_name (str | None)
conley_lag_cutoff (int | None)
spillover_effects (DataFrame | None)
ring_breakpoints (List[float] | None)
d_bar (float | None)
n_units_ever_in_ring (Dict[str, int] | None)
n_far_away_obs (int | None)
is_staggered (bool | None)
event_study (bool | None)
stage1_n_obs (int | None)
anticipation (int | None)
att_dynamic (DataFrame | None)
event_study_effects (Dict[int, Dict[str, Any]] | None)
horizon_max (int | None)
reference_period (int | None)
n_psu (int | None)
n_strata (int | None)

Return type:

None

classmethod __new__(*args, **kwargs)#