term_drop_diagnostics

SuperGLM.term_drop_diagnostics(

X: object,

y: NDArray,

sample_weight: NDArray | None = None,

offset: NDArray | None = None,

*,

mode: str = 'refit',

X_val: object | None = None,

y_val: NDArray | None = None,

sample_weight_val: NDArray | None = None,

offset_val: NDArray | None = None,

) → DataFrame

Drop-term diagnostics: AIC/BIC deltas or holdout loss deltas.

Parameters:

X, ytraining rows and response

Used for refit mode and as the identity anchor for same-object holdout fallback.

sample_weight, offsetarray-like, optional

Training/refit weights and offset.

mode{“refit”, “holdout”}

"refit" calls drop1() and adds delta IC columns. "holdout" zeros each term on a validation set (no refit).

X_val, y_valoptional

Validation data for mode="holdout".

sample_weight_val, offset_valarray-like, optional

Validation-specific geometry for holdout mode. Training vectors are reused only when X_val is X and y_val is y; separate validation objects require sample_weight_val when sample_weight is supplied and require offset_val for an offset-fitted model.