Plotting¶

rgram.plot_diagnostics(
    estimator,
    X,
    y,
)

Plot residuals and observed-versus-predicted values without sorting rows.

Parameters¶

estimator Regressogram or KernelSmoother

Successfully fitted model providing regression_diagnostics. For a pipeline, use the final estimator with transformed feature coordinates.

X array-like of shape (n_samples,) or (n_samples, 1)

Evaluation feature values, aligned with y.

y array-like of shape (n_samples,) or (n_samples, 1)

Observed responses, for training or held-out observations.

Returns¶

figure matplotlib.figure.Figure

Figure containing both diagnostic panels and a support-count title.

axes ndarray of shape (2,)

Residual-versus-feature axis and observed-versus-predicted axis.

Raises¶

ImportError

If Matplotlib is absent; install rgram[plot].

sklearn.exceptions.NotFittedError

If the estimator has no successful fit.

ValueError

If evaluation pairs are invalid or strict prediction policies raise.

Notes¶

Uses scatter points, never connecting or sorting observation rows. Unsupported rows remain in the diagnostic table and are counted in the title, though non-finite points cannot be drawn. No data is modified. The caller chooses whether to display, save, or close the returned figure.

Examples¶

>>> from rgram import Regressogram, plot_diagnostics
>>> model = Regressogram(n_bins=2).fit([0., 1., 2.], [1., 3., 4.])
>>> figure, axes = plot_diagnostics(model, [0., 1., 2.], [1., 3., 4.])
>>> len(axes)
2