Lattice implementation of the recursive maximum likelihood algorithm
B. Friedlander, Lennart Ljung, Martin E. Morf · 1981
A lattice implementation of the extended least-squares algorithm was recently developed. The convergence of that algorithm depends on a positive real condition, a fact which limits its usefulness in some applications. In this note the lattice equivalent of the recursive maximum likelihood algorithm is derived. This technique does not require the positive real condition and is suitable for modeling and prediction of general ARMA time-series.