plot_diagnostics

SuperGLM.plot_diagnostics(

X: object,

y: NDArray,

sample_weight: NDArray | None = None,

offset: NDArray | None = None,

*,

n_sim: int = 100,

figsize: tuple[float, float] | None = None,

max_points: int = 50000,

seed: int = 42,

residual_type: str = 'auto',

)

GLM/GAM diagnostic figure with simulation-based Q-Q envelope.

Four panels using quantile residuals (Dunn & Smyth 1996):

1. Q-Q with simulation envelope

2. Calibration (sample-weighted observed vs predicted)

3. Residuals vs Linear Predictor

4. Residual distribution (histogram + N(0,1) overlay)

Parameters:

Xpandas or eager Polars DataFrame

Design matrix.

yNDArray

Response vector.

sample_weightNDArray or None

Observation weights, read under the model’s declared weight_semantics: replication counts under "frequency", precisions under "prior". The quantile residuals below take the row’s own marginal distribution from that contract, so a prior weight enters the reference distribution and a replication count does not.

offsetNDArray or None

Optional offset.

n_simint

Number of simulation replicates for the Q-Q envelope.

figsizetuple or None

Figure size in inches. Defaults to (10, 8).

max_pointsint

Threshold for scatter vs hexbin rendering.

seedint

Random seed for quantile residuals, simulation, and subsampling.

residual_typestr

Deprecated since version Ignored.: All panels use quantile residuals.

Returns:

matplotlib.figure.Figure

A figure with 4 diagnostic subplots.