DGP Simulation

simulation_step(
    name: str = "datagen",
    *,
    T: int,
    shocks: Mapping[str, Shock | Callable | ndarray] | None = None,
    seed_increment: int | Literal["auto"] = "auto",
    shock_scale: float = 1.0,
    x0: ndarray | None = None,
    observables: bool = True,
) -> MCStep

simulation_step wraps simulate_dgp(...). It requires MCPipeline.run(..., dgp=...) and calls dgp.sim(...) in each replication.

Inputs:

Name	Description
T	Number of simulated periods, excluding the initial state.
shocks	Shock specification passed into DGP simulation after per-replication seed handling.
seed_increment	Integer seed offset per replication, or "auto" to increment by the number of seeded Shock objects.
shock_scale	Shock scaling passed into SolvedModel.sim(...).
x0	Optional initial state.
observables	If True, observable paths are included in MCData.observables.

Seed Convention

For generator-style Shock objects with integer seeds, replication rep_idx receives shock.seed + rep_idx * seed_increment. With seed_increment="auto", the increment is the number of seeded Shock entries. Array shocks and callable shocks are passed through unchanged.