posterior_predictive

SuperLSS.posterior_predictive(

X: object | EagerFrame,

n_draws: int = 200,

*,

sample_weight: NDArray | None = None,

**kwargs: Any,

) → NDArray[float64]

Simulate responses for the rows of X through the family’s quantile.

With parameter_uncertainty=True each draw uses its own coefficient draw, so the spread is predictive and not merely conditional; reduce collapses each block of draws before it is materialised, which is what makes a book-level total affordable. sample_weight enters as the row’s prior weight, and under frequency semantics the call is refused rather than simulating a law the weight does not describe.