NonlinearSSM.extended_smoother

NonlinearSSM.extended_smoother(y)

The extended Rauch-Tung-Striebel smoother.

Runs the extended filter, then Rauch’s backward recursion with the transition Jacobian evaluated at each filtered mean. The linearization error therefore compounds through both passes: exact when f and g are linear, degrading faster than the forward filter as curvature grows. When this and the unscented smoother disagree materially, trust the unscented one.

Parameters:

y (ArrayLike) – Data, (n, p); numpy.nan rows are missing.

Return type:

_RtsSmootherResult

Returns:

The _RtsSmootherResult with method="extended".

Raises:

• SpecificationError – If the model carries custom hooks.

• NumericalError – If either pass degenerates.