LinearGaussianSSM._from_arma#

classmethod LinearGaussianSSM._from_arma(phi_star, theta_star, sigma2, obs_intercept)[source]#

Build the Harvey state-space form of an ARMA process.

The state dimension is max(p, q + 1), which is the minimal realization: a larger companion would be observationally equivalent but would make the Kalman recursion carry redundant states.

Parameters:

  • phi_star (NDArray[double]) – Expanded AR coefficients.

  • theta_star (NDArray[double]) – Expanded MA coefficients.

  • sigma2 (float) – Innovation variance.

  • obs_intercept (NDArray[double]) – Per-observation mean shift from the regression block.

Return type:

_LinearGaussianStateSpace

Returns:

The configured _LinearGaussianStateSpace.