Results

class hddid.HDDIDResult(parametric_estimates=<factory>, nonparametric_estimates=<factory>, standard_errors=<factory>, intervals=<factory>, diagnostics=None)

User-facing container for HDDID estimation results.

Parameters:

• parametric_estimates (dict[str, Any])

• nonparametric_estimates (dict[str, Any])

• standard_errors (dict[str, Any])

• intervals (dict[str, ConfidenceInterval])

• diagnostics (ResultDiagnostics | None)

parametric_estimates

Named parametric point estimates (beta_hat, t_hat).

Type:

dict

nonparametric_estimates

Named nonparametric estimates (gamma_hat, f_hat_at_z0).

Type:

dict

standard_errors

Standard errors keyed by estimate name.

Type:

dict

intervals

Confidence intervals keyed by estimate name.

Type:

dict of ConfidenceInterval

diagnostics

Cross-fitting and solver diagnostics.

Type:

ResultDiagnostics or None

class hddid.ConfidenceInterval(lower, upper, level=None)

Pointwise confidence interval for a scalar or vector estimate.

Parameters:

• lower (float | ndarray)

• upper (float | ndarray)

• level (float | None)

lower

Lower confidence bound.

Type:

float or ndarray

upper

Upper confidence bound.

Type:

float or ndarray

level

Nominal coverage level (e.g. 0.95).

Type:

float or None

class hddid.UniformBand(lower, upper, level=None, critical_value=None, n_boot=None, random_state=None)

Uniform confidence band for the nonparametric function.

Parameters:

• lower (float | ndarray)

• upper (float | ndarray)

• level (float | None)

• critical_value (float | None)

• n_boot (int | None)

• random_state (int | None)

lower

Lower band boundary.

Type:

float or ndarray

upper

Upper band boundary.

Type:

float or ndarray

level

Nominal coverage level.

Type:

float or None

critical_value

Bootstrap critical value for uniform coverage.

Type:

float or None

n_boot

Number of bootstrap replications used.

Type:

int or None

random_state

Random seed for bootstrap reproducibility.

Type:

int or None

Diagnostics

class hddid.FoldDiagnostics(fold_id, basis_family=None, basis_degree=None, oracle_lane=None, n_holdout_raw=None, n_trimmed_propensity=None, n_valid_holdout=None, trim_lower=None, trim_upper=None)

Per-fold diagnostic summary from cross-fitting.

Parameters:

• fold_id (int)

• basis_family (str | None)

• basis_degree (int | None)

• oracle_lane (str | None)

• n_holdout_raw (int | None)

• n_trimmed_propensity (int | None)

• n_valid_holdout (int | None)

• trim_lower (float | None)

• trim_upper (float | None)

fold_id

One-based fold index.

Type:

int

basis_family

Sieve basis family used in this fold.

Type:

str or None

basis_degree

Sieve basis degree.

Type:

int or None

oracle_lane

Oracle nuisance lane identifier (simulation only).

Type:

str or None

n_holdout_raw

Total holdout observations before trimming.

Type:

int or None

n_trimmed_propensity

Observations removed by propensity score trimming.

Type:

int or None

n_valid_holdout

Observations remaining after trimming.

Type:

int or None

trim_lower

Lower propensity trimming threshold applied.

Type:

float or None

trim_upper

Upper propensity trimming threshold applied.

Type:

float or None

class hddid.ResultDiagnostics(basis_family, basis_degree, oracle_lane, fold_diagnostics=<factory>, n_holdout_raw=None, n_trimmed_propensity=None, n_valid_holdout=None, trim_lower=None, trim_upper=None, optimization_metadata=<factory>)

Aggregated diagnostics across all cross-fitting folds.

Parameters:

• basis_family (str)

• basis_degree (int)

• oracle_lane (str)

• fold_diagnostics (list[FoldDiagnostics])

• n_holdout_raw (int | None)

• n_trimmed_propensity (int | None)

• n_valid_holdout (int | None)

• trim_lower (float | None)

• trim_upper (float | None)

• optimization_metadata (dict[str, Any])

basis_family

Sieve basis family.

Type:

str

basis_degree

Sieve basis degree.

Type:

int

oracle_lane

Oracle lane identifier.

Type:

str

fold_diagnostics

Per-fold diagnostic records.

Type:

list of FoldDiagnostics

n_holdout_raw

Total holdout observations (all folds combined).

Type:

int or None

n_trimmed_propensity

Total observations trimmed across folds.

Type:

int or None

n_valid_holdout

Total valid observations after trimming.

Type:

int or None

trim_lower

Lower propensity trimming threshold.

Type:

float or None

trim_upper

Upper propensity trimming threshold.

Type:

float or None

optimization_metadata

Solver convergence and iteration metadata.

Type:

dict