|
|
|
|
| import numpy as np
|
| from scipy.stats import norm, t
|
|
|
| from .estimators import estimate_mean, estimate_sigma
|
|
|
|
|
| def ci_mean_analytic(
|
| *,
|
| data,
|
| estimator,
|
| alpha,
|
| dist,
|
| sigma_estimator,
|
| trim_param=None,
|
| winsor_limits=None,
|
| weights=None,
|
| ):
|
| """
|
| Analytic confidence interval for the mean.
|
|
|
| - Mean is computed with the user-chosen mean estimator.
|
| - σ is computed with the user-chosen deviation estimator.
|
| """
|
| n = len(data)
|
|
|
| mu_hat = estimate_mean(
|
| data,
|
| estimator,
|
| trim_param=trim_param,
|
| winsor_limits=winsor_limits,
|
| weights=weights,
|
| )
|
|
|
| sigma_hat = estimate_sigma(
|
| data=data,
|
| estimator=sigma_estimator,
|
| )
|
|
|
| scale = sigma_hat / np.sqrt(n)
|
|
|
| if dist == "t":
|
| return (
|
| t.ppf(alpha / 2, n - 1, loc=mu_hat, scale=scale),
|
| t.ppf(1 - alpha / 2, n - 1, loc=mu_hat, scale=scale),
|
| )
|
|
|
| return (
|
| norm.ppf(alpha / 2, loc=mu_hat, scale=scale),
|
| norm.ppf(1 - alpha / 2, loc=mu_hat, scale=scale),
|
| )
|
|
|
|
|
| def ci_mean_bootstrap(
|
| *,
|
| data,
|
| estimator,
|
| alpha,
|
| B,
|
| trim_param=None,
|
| winsor_limits=None,
|
| weights=None,
|
| ):
|
| """
|
| Bootstrap CI for the mean using the user-chosen mean estimator.
|
| """
|
| data = np.asarray(data)
|
| n = len(data)
|
|
|
| boot_stats = np.empty(B)
|
|
|
| for b in range(B):
|
| idx = np.random.choice(n, size=n, replace=True)
|
| boot_data = data[idx]
|
|
|
| boot_weights = None
|
| if weights is not None:
|
| boot_weights = np.asarray(weights)[idx]
|
|
|
| boot_stats[b] = estimate_mean(
|
| boot_data,
|
| estimator,
|
| trim_param=trim_param,
|
| winsor_limits=winsor_limits,
|
| weights=boot_weights,
|
| )
|
|
|
| return np.quantile(boot_stats, [alpha / 2, 1 - alpha / 2])
|
|
|
