term_inference#

SuperLSS.term_inference(
parameter: str,
term: str,
*,
covariance: Literal['fixed', 'corrected'] = 'fixed',
X_train: object | EagerFrame | None = None,
**kwargs: Any,
) → ParameterTermEffect#

Return one term of one parameter on that parameter’s link scale.

The term is swept over its training range with every other covariate held at its training centre, and carries the Bayesian pointwise band of Marra and Wood (2012) beside the max-deviation simultaneous band of Ruppert, Wand and Carroll (2003). The sweep needs the frame the model was fitted on: it is kept by reference at fit/fit_reml but does not survive serialization, so a restored model must be given X_train=.