UCSVPosterior.forecast#

UCSVPosterior.forecast(steps, *, n_paths=2000, seed=None)[source]#

Predictive draws of the series, (n_paths, steps).

Each path picks a kept draw, then propagates both random-walk log variances and the trend forward and adds the irregular.

Parameters:

  • steps (int) – Horizons ahead, at least 1.

  • n_paths (int) – Predictive paths.

  • seed (int | None) – Seed.

Raises:

SpecificationError – If steps is not positive.

Return type:

NDArray[double]