smooth.OMG.multicov

OMG.multicov(type='analytical', h=10, nsim=1000)

Multi-step forecast-error covariance — not defined for OMG.

R’s multicov is dispatched on adam / smooth classes via the state-space matrices (F, W, g, σ²). OMG has two parallel sub-models joined by the non-linear link p = pA / (pA + pB); the resulting joint multi-step distribution has no closed-form covariance from those matrices and R does not define a multicov.omg method either (it would dispatch to multicov.smooth and crash on the missing $persistence).

Call smooth.adam_general.core.adam.ADAM.multicov() on each sub-model — model.model_a.multicov(...) / model.model_b.multicov(...) — for per-side multi-step covariance instead. Raises NotImplementedError.


Parent Class: OMG