Probabilistic and interval regression¶

This page explains the two regression modes and how they differ.

For full guarantees, assumptions, and non-guarantees, use Calibrated interval semantics.

Mode routing¶

Mode	API signal	Output	Primary question
Percentile or interval regression	`predict(..., uq_interval=True, low_high_percentiles=...)`	Numeric prediction with CPS percentile interval	Where will `y` fall?
Probabilistic or thresholded regression	`predict_proba(..., threshold=...)`	Calibrated event probability with interval	How likely is an event on `y`?

Examples¶

pred, (low, high) = explainer.predict(
    X,
    uq_interval=True,
    low_high_percentiles=(5, 95),
)

p, (plo, phi) = explainer.predict_proba(
    X,
    threshold=150,
    uq_interval=True,
)

Interval mode and probabilistic mode can be used on the same calibrated explainer. The threshold argument selects probabilistic mode.