estimate_nonlinear_did_for_panel

nonlinear_did.estimate_nonlinear_did_for_panel(
    ds: pl.DataFrame,
    outcome: ColumnName,
    treatment_group: ColumnName | TrtGroupConfig,
    time: ColumnName,
    unit: ColumnName,
    qs: ArrayLike = MEDIAN,
    *,
    weights: ColumnName | None = None,
    n_anticipation_periods: int = 0,
    base_period: BasePeriod = BasePeriod.UNIVERSAL,
    control_group: ControlGroup = ControlGroup.NEVER_TREATED,
    counterfactual_model: CounterfactualModel = CounterfactualModel.CIC,
    bootstrap_config: BootstrapConfig | int = 1000,
)

Estimate nonlinear difference-in-differences effects for panel data.

Implements the changes-in-changes (CiC) and quantile difference-in-differences (QDiD) estimators with the aggregations of Callaway and Li. Returns the group, event-study, overall and group-time effects.

Parameters

Name	Type	Description	Default
ds	`polars`.`DataFrame`	Panel data in long format.	required
outcome	`ColumnName`	Name of the outcome column.	required
treatment_group	ColumnName or `TrtGroupConfig`	Column holding the first treatment period of a unit. A plain string treats the column as already coded with `inf` for never-treated units; a `TrtGroupConfig` maps a custom never-treated value to `inf`.	required
time	`ColumnName`	Name of the time period column.	required
unit	`ColumnName`	Name of the unit identifier column.	required
qs	`array_like`	Quantiles in `(0, 1)` at which to estimate the effects.	`0.5`
weights	`ColumnName`	Name of a column with sampling weights.	`None`
n_anticipation_periods	`int`	Number of periods before treatment in which units may anticipate it.	`0`
base_period	`BasePeriod`	Pre-treatment period used for the comparisons.	`BasePeriod.UNIVERSAL`
control_group	`ControlGroup`	Comparison group used for the counterfactual.	`ControlGroup.NEVER_TREATED`
counterfactual_model	`CounterfactualModel`	Model used to construct the counterfactual distribution.	`CounterfactualModel.CIC`
bootstrap_config	`BootstrapConfig` or int	Bootstrap settings, or the number of replications.	`1000`

Returns

Name	Type	Description
	`NonlinearDidResults`	The group, event-study, overall and group-time effects.