Abstract
The extension of Kalman filtering to the case of unknown variances is known as adaptive filtering. Duncan and Horn's (1972) representation of the Kalman time-series model is exploited to derive variance-component estimates from the standpoint of regression analysis. Specifically, the MINQUE of Rao (1970) and the AUE of Horn, Horn, and Duncan (1975) are applied. These estimates are then used to formulate adaptive filters. The adaptive filters are applied to simulated data, and their performance is compared to the method of Carew and Belanger (1973).