Skip to content
EvalSuite
Documentation menu

Calibration

Calibration intercept

Implementedcalibration.calibration_intercept

Definition

Calibration-in-the-large: intercept a of logit P(y=1) = a + logit(p) with the slope fixed at 1. 0 is ideal; negative means risks are overestimated on average, positive underestimated.

Formula

logit P(y=1) = a + logit(p̂) (offset)

Range: (−∞, ∞), ideal 0

Inputs and outputs

  • y_true: binary outcomes
  • y_prob: predicted risks in (0, 1)

Returns: MetricResult

Assumptions

No assumptions beyond valid, aligned inputs of the documented types.

Limitations

No metric-specific limitations are documented yet. Interpret the value alongside the task, data, and other metrics.

Python API

PythonSince v0.2.0
import evalsuite as es

es.calibration_intercept(y_true, y_prob)

References

  1. Cox DR. Two further applications of a model for binary regression. Biometrika. 1958;45(3-4):562-565.
  2. Van Calster B, McLernon DJ, van Smeden M, Wynants L, Steyerberg EW. Calibration: the Achilles heel of predictive analytics. BMC Med. 2019;17(1):230.

Implementation status