API reference¶
API stability¶
Rgram has an unstable API. Parameters, methods, and result schemas may change between releases without backward compatibility or a deprecation period. Pin the package version for reproducible analyses.
All top-level exports are documented below. The estimators also inherit
scikit-learn parameter, scoring, and metadata-request methods. Metadata-request
methods are included for completeness; their presence does not mean Rgram
implements full metadata routing, especially through CoverageSearchCV pipelines.
Use the explicit weight forwarding described in the sklearn guide.
Public interface¶
| Object | Purpose |
|---|---|
| Regressogram | Binned response summaries. |
| KernelSmoother | Local-constant and local-linear kernel regression. |
| CoverageSearchCV | Validation MSE search requiring explicit prediction coverage. |
| rgram.array_aggregation | Numeric per-bin scalar callback adapter. |
| rgram.weighted_aggregation | Native Polars weighted callback adapter. |
| rgram.quantile | Explicit unweighted quantile reducer. |
| plot_diagnostics | Optional residual and observed-versus-predicted figure. |
| Warning categories | Filter support, ordering, extrapolation, binning, and numerical advisories. |
The helper factories return cloneable classes documented under
rgram.aggregation. Module internals are explained separately in the
implementation guide.