dualbounds.utilities.compute_est_bounds¶

dualbounds.utilities.compute_est_bounds(summands: array, clusters: callable | None = None, func: callable | None = None, B: int = 1000, alpha: float = 0.05)[source]¶

Helper to computes confidence intervals.

Specifically, computes intervals for

func(summands[k].mean(axis=-1))

Provides lower CI for k=0 and upper CI for k=1.

Parameters:¶

summands : np.array¶

(2, n)-shaped array or (2, n, d)-shaped array

func : callable¶

A callable that maps a 1D np.array to a scalar. Defaults to func=lambda x: x.mean().

clusters : np.array¶

n-shaped array where clusters[i] = j means that observation i is in the jth cluster.

B : int¶

Number of bootstrap draws to use. Ignored unless func or clusters are provided.

alpha : float¶

Nominal level.

Returns:¶

• ests (np.array) – 2-shaped array of lower and upper estimators (sample mean).

• ses (np.array) – 2-shaped array of standard errors.

• bounds (np.array) – 2-shaped array of lower/upper confidence bounds.

Notes

This is meant primarily for internal use in the DualBounds class.