Honest calibration assessment for binary outcome predictions


METADATA ONLY
Loading...

Date

2023-09

Publication Type

Journal Article

ETH Bibliography

yes

Citations

Altmetric
METADATA ONLY

Data

Rights / License

Abstract

Probability predictions from binary regressions or machine learning methods ought to be calibrated: if an event is predicted to occur with probability x, it should materialize with approximately that frequency, which means that the so-called calibration curvep(middot) should equal the identity, i.e., p(x) = x for all x in the unit interval. We propose honest calibration assessment based on novel confidence bands for the calibration curve, which are valid subject to only the natural assumption of isotonicity. Besides testing the classical goodness-of-fit null hypothesis of perfect calibration, our bands facilitate inverted goodness-of-fit tests whose rejection allows for the sought-after conclusion of a sufficiently well-specified model. We show that our bands have a finite-sample coverage guarantee, are narrower than those of existing approaches, and adapt to the local smoothness of the calibration curve p and the local variance of the binary observations. In an application to modelling predictions of an infant having low birth weight, the bounds give informative insights into model calibration.

Permanent link

Publication status

published

Editor

Book title

Journal / series

Volume

110 (3)

Pages / Article No.

663 - 680

Publisher

Oxford University Press

Event

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Binary regression; Calibration validation; Confidence band; Goodness-of-fit; Isotonic regression; Universally valid inference

Organisational unit

Notes

Funding

Related publications and datasets