MarkovSwitchingAR.update_transition#

static MarkovSwitchingAR.update_transition(smoothed, joint, floor)[source]#

M-step transition update from expected transition counts.

Probabilities are floored before renormalizing, so a regime that becomes momentarily unvisited can still be re-entered instead of being absorbed.

Parameters:

  • smoothed (NDArray[double]) – Smoothed regime probabilities.

  • joint (NDArray[double]) – Smoothed joint probabilities Pr(S_t = i, S_{t+1} = j | y).

  • floor (float) – Minimum admissible probability.

Return type:

NDArray[double]

Returns:

The updated row-stochastic transition matrix.