📉  Quantile Forecasts

Quantile Forecasts#

Forecasts of specific quantiles of the outcome distribution.

Important

✨ AI-generated content. This page was written with the assistance of an AI language model and is provided as a learning aid. Despite careful review, it may still contain mistakes, omissions, or out-of-date information. Whether you are new to the topic, a team lead, or a senior practitioner, treat it as a starting point rather than an authoritative reference: read it critically and independently verify anything you act on (code, commands, figures, and factual claims) against official documentation and primary sources before relying on it.

What it is#

A quantile forecast expresses the prediction not as one number but as one or more conditional quantiles of the future value’s distribution — for example the 10th, 50th and 90th percentiles. The 0.5 level is the median; a pair like 0.1 / 0.9 brackets the outcome, saying it is unlikely (\(\le\) 10%) to fall below the lower or above the upper. Formally the \(\tau\)-quantile forecast satisfies

\[\Pr\!\left(Y \le \hat{y}_\tau \mid X\right) = \tau.\]

Why use them#

They convey uncertainty and asymmetry directly: you can act differently when under- and over-shooting carry different costs, without assuming a parametric (e.g. Gaussian) shape for the distribution.

How they’re scored#

Quantile forecasts are evaluated with the pinball (quantile) loss, matched to each level; the CRPS (Continuous Ranked Probability Score) generalizes it by integrating the pinball loss across all quantiles, giving a single distributional score.


Theme: Risk & Probabilistic Forecasting  ·  All terminology



See also

Source article Adapted (context, re-expressed) in our own words from: Quantile Forecasts (insightful-data-lab.com).

Tags: purpose: reference topic: terminology level: advanced