THE USE OF BIASED LEAST-SQUARES ESTIMATORS FOR PARAMETERS IN DISCRETE-TIME PULSE-RESPONSE MODELS

The use of biased least-squares methods for the estimation of pulse-response coefficients in discrete-time process models is proposed. The two specific approaches considered here are partial least squares (PLS) and a method based on the singular value decomposition (SVD). Example applications include the estimation of the open-loop impulse and step responses of a pilot-scale anaerobic wastewater treatment process. The SVD and PLS methods give very similar results for the problems studied. The PLS method typically requires much less computer time, however.