causalis.dgp.panel_data_did.base¶
Module Contents¶
Classes¶
Low-level generator for realistic simultaneous-adoption DID panels. |
API¶
- class causalis.dgp.panel_data_did.base.PanelDIDGeneratorConfig¶
- n_treated_units: int¶
20
- n_control_units: int¶
60
- n_pre_periods: int¶
24
- n_post_periods: int¶
12
- time_start: int¶
1
- time_freq: str¶
‘M’
- calendar_start: str¶
‘2021-01’
- treated_prefix: str¶
‘treated_market_’
- control_prefix: str¶
‘control_market_’
- random_state: Optional[int]¶
42
- return_panel_data: bool¶
True
- outcome_distribution: Literal[gaussian, gamma, poisson]¶
‘gaussian’
- gamma_shape: float¶
9.0
- exposure_log_mean: float¶
7.1
- exposure_log_std: float¶
0.45
- treated_selection_shift: float¶
0.12
- common_factor_std_log: float¶
0.05
- unit_noise_std_log: float¶
0.07
- outcome_noise_std_log: float¶
0.06
- gaussian_noise_std: float¶
90.0
- rho_common: float¶
0.45
- rho_unit: float¶
0.35
- seasonality_strength: float¶
0.16
- parallel_trend_violation: float¶
0.0
- treatment_effect_rate: float¶
0.08
- treatment_effect_slope: float¶
0.0
- treatment_effect_heterogeneity_std: float¶
0.025
- class causalis.dgp.panel_data_did.base.PanelDIDGenerator(config: causalis.dgp.panel_data_did.base.PanelDIDGeneratorConfig)¶
Low-level generator for realistic simultaneous-adoption DID panels.
Initialization
- covariate_cols: tuple[str, ...]¶
(‘exposure’, ‘avg_order_value’, ‘market_competition’, ‘macro_index’, ‘seasonality_index’)
- cluster_col: str¶
‘region’
- generate(*, return_panel_data: Optional[bool] = None) Union[pandas.DataFrame, causalis.data_contracts.panel_data_did.PanelDataDID]¶