A simple algorithm to factorize the autocovariance function of moving average process chains (Q1966362)

This paper presents a simple recursive method to find the moving average coefficients from autocovariances, which uses standard matrix operations like addition, multiplication, inversion. Such a method is easier to implementation than the Kalman filter or a procedure based on finding zeros of autocovariance generating functions. Several relationships between the moving average coefficients and the roots of the relevant polynomials are derived. Both theoretical foundations and details of practical implementation of the recursive algorithm are presented, as well as comparisons of the computing times of the standard procedure and the proposed algorithm.