Prediction Interval & Uncertainty Visualizations

[kiplangatkorir]

Problem

Uncertainty estimates are difficult to interpret without visual context. Users need clear, standardized visualizations to assess calibration, sharpness, and reliability of prediction intervals.

Proposal

Add a set of first-class visualization utilities for uncertainty outputs, usable directly in notebooks and scripts.

Scope

• Prediction interval plots for regression
• Confidence bands over input domain or time
• Point predictions with uncertainty overlays
• Compatibility with:
  - Bootstrap
  - Conformal prediction
  - Ensembles
• Minimal dependencies (matplotlib preferred)

API Sketch

from ueq.visualization import plot_intervals

plot_intervals(
    x,
    y_pred,
    lower,
    upper
)

Acceptance Criteria

• Works out-of-the-box for common UQ outputs
• Clear labeling of nominal vs empirical coverage
• Publication-quality defaults
• Notebook and script compatibility
• Example gallery in documentation

[refexa]

Issue resolved, check and feedback please.

[kiplangatkorir]

Great @refexa Open a PR so that I may review it, give feedback and possibly merge the PR