MarkovSwitchingSSM.effective_sample#

MarkovSwitchingSSM.effective_sample(y)[source]#

Split a series into the modelled target and its lag block.

Parameters:

y (ArrayLike) – The untrimmed series.

Return type:

tuple[NDArray[double], NDArray[double]]

Returns:

A tuple (target, lags); target drops the first order observations and lags is the (T_eff, p) matrix the conditional means are built from.

Raises:

DimensionError – If the series is not 1-D or is shorter than the autoregressive order.