A QR-type algorithm for fitting the delta AR model to autocorrelation windowed data

A QR algorithm is developed to fit the delta autoregressive (DAR) model of Vijayan et al. (see IEEE Trans. Automat. Cont. vol.36, p.314, 1991) to autocorrelation windowed sampled data. To obtain the DAR model parameters, one must solve a linear system in the matrix called Q/sub n/. Unfortunately, there is presently no way to obtain good estimates of Q/sub n/. The proposed QR-type algorithm overcomes this problem by computing the DAR model parameters without the need for estimating Q/sub n/ directly. The QR algorithm proposed is a simple modification of the classical QR algorithm for the conventional AR model due to C.P. Rialan and L.L. Scharf (1986).<
>