MarkovSwitchingSSM

cultivars.state_space.regime_switching.MarkovSwitchingSSM

alias of _MarkovSwitchingStateSpace

Attributes

ar_params	Per-regime autoregressive coefficients, shape (K, p).
initial_prob	Distribution of the regime at the first modelled observation.
intercepts	Per-regime intercepts.
k_endog	Observed dimension; one, since the chain drives a scalar series.
k_states	the number of regimes.
n_regimes	The number of regimes, named as the chain rather than as a dimension.
order	Autoregressive order of the per-regime observation equation.
transition	The row-stochastic regime transition matrix.
variances	Per-regime innovation variances.

Methods

conditional_means(lags)	Per-regime conditional means E[y_t | S_t = j, lags].
density_matrix(target, lags)	The (T_eff, K) log-density matrix the recursions consume.
effective_sample(y)	Split a series into the modelled target and its lag block.
filter(y)	Filter a series, returning regime probabilities and the likelihood.
filter_densities(log_density)	Run the Hamilton filter over an already-built density matrix.
log_densities(target, means)	Gaussian log conditional densities, one column per regime.
loglikelihood(y)	The exact log-likelihood of a series under this system.
smooth(y)	Filter, then smooth, returning full-sample regime probabilities.
smooth_densities(filtered)	Run the Kim smoother over a completed forward pass.