Modify the code for Gaussian elimination to exploit the structure of a banded matrix. Store the matrix taking n(p + q + 1) locations. For the covariance matrix AN+n of an ARMA(p, q) process,...

Modify the code for Gaussian elimination to exploit the structure of a banded matrix. Store the matrix taking n(p + q + 1) locations.

For the covariance matrix AN+n of an ARMA(p, q) process, partitioned as

                    AN+n =  AN A12 A21 An  , N n

give an algorithm for computing the forecast covariance matrix, An − A21A−1 N A12.