causalis.dgp.causaldata.preperiod

Module Contents

Classes

PreCorrSpec

Functions

corr_on_scale
calibrate_sigma_for_target_corr	Find sigma such that Corr(T(y_pre_base + sigma*eps), T(y_post)) ~ target_corr. Returns (sigma, achieved_corr).
add_preperiod_covariate	Standardized utility to add a calibrated pre-period covariate to a DataFrame.

Data

CorrMethod
Transform

API

causalis.dgp.causaldata.preperiod.CorrMethod

None

causalis.dgp.causaldata.preperiod.Transform

None

causalis.dgp.causaldata.preperiod.corr_on_scale(y_pre: numpy.ndarray, y_post: numpy.ndarray, *, transform: causalis.dgp.causaldata.preperiod.Transform = 'log1p', winsor_q: Optional[float] = 0.999, method: causalis.dgp.causaldata.preperiod.CorrMethod = 'pearson') → float

class causalis.dgp.causaldata.preperiod.PreCorrSpec

target_corr: float

0.7

transform: causalis.dgp.causaldata.preperiod.Transform

‘log1p’

winsor_q: Optional[float]

0.999

method: causalis.dgp.causaldata.preperiod.CorrMethod

‘pearson’

sigma_lo: float

0.0

sigma_hi: float

50.0

sigma_tol: float

0.001

max_iter: int

40

causalis.dgp.causaldata.preperiod.calibrate_sigma_for_target_corr(y_pre_base: numpy.ndarray, y_post: numpy.ndarray, rng: numpy.random.Generator, spec: causalis.dgp.causaldata.preperiod.PreCorrSpec, *, noise: Optional[numpy.ndarray] = None) → Tuple[float, float]

Find sigma such that Corr(T(y_pre_base + sigma*eps), T(y_post)) ~ target_corr. Returns (sigma, achieved_corr).

causalis.dgp.causaldata.preperiod.add_preperiod_covariate(df: pandas.DataFrame, y_col: str, d_col: str, pre_name: str, base_builder: Callable[[pandas.DataFrame], numpy.ndarray], spec: causalis.dgp.causaldata.preperiod.PreCorrSpec, rng: numpy.random.Generator, mask: Optional[numpy.ndarray] = None) → pandas.DataFrame

Standardized utility to add a calibrated pre-period covariate to a DataFrame.

Parameters

df : pd.DataFrame The dataset. y_col : str Name of the outcome column. d_col : str Name of the treatment column. pre_name : str Name of the new pre-period covariate column. base_builder : callable Function df -> y_pre_base (np.ndarray) providing the shared signal. spec : PreCorrSpec Specification for target correlation and scale. rng : np.random.Generator Random number generator. mask : np.ndarray, optional Boolean mask of rows to use for calibration (e.g. control group). If None, use control group (d == 0).