Corrections

This module provides correction functions for multiple hypothesis testing. To be used when building a calibrator for multivariate regression or object detection.

corrections.bonferroni(alpha, nvars=1)

Bonferroni correction for multiple comparisons.

Parameters:

  • alpha (float) – nominal coverage level.

  • nvars (int) – number of output features.

Returns:

corrected coverage level.

Return type:

float or ndarray.

corrections.weighted_bonferroni(alpha, weights)

Weighted Bonferroni correction for multiple comparisons.

Parameters:

  • alpha (float) – nominal coverage level.

  • weights (np.ndarray) – weights associated to each output feature.

Returns:

array of corrected featurewise coverage levels.

Return type:

np.ndarray.