residual_set

SuperLSS.residual_set(

X: object | EagerFrame,

y: NDArray,

*,

kind: Literal['pit', 'quantile'] = 'quantile',

sample_weight: NDArray | None = None,

offsets: Mapping[str, NDArray] | None = None,

seed: int = 42,

) → ResidualSet

Return the full residual payload of y under the fitted parameters.

The payload carries the probability-integral transform of Dunn and Smyth (1996) and its normal inverse together with the parameters, the response and the weights they were read under, which is what the Q-Q, worm, PIT and binned checks all consume. sample_weight is read under the model’s declared contract: a prior weight is part of each row’s law, so the transform is the prior-weighted one, while a frequency weight is replication that the checks expand.