User guide¶

Conformity separates fitting a point model from calibrating its uncertainty. Start with the lifecycle, choose a region suited to your target, then evaluate coverage and usefulness on observations used by neither step.

1. Estimators and pipelines: cloning, fitting, calibration, preprocessing placement and model selection.
2. Calibration and guarantees: held-out splits, exchangeability, finite-sample thresholds and the limits of marginal coverage.
3. Regression intervals: symmetric or asymmetric residual corrections, marginal coverage and the tradeoff between coverage and width.
4. Classification sets: LAC and deterministic APS, class ordering, p-values and ambiguous predictions.
5. Diagnostics and errors: coverage, efficiency, weights, failure modes and independent evaluation.
6. Theory: score definitions, order statistics and the rank proof.
7. Plotting: visualize intervals and investigate missed outcomes.

For complete workflows, visit the examples gallery. The API reference describes callable contracts and fitted state.