Why Yule-Walker should not be used for autoregressive modelling

Autoregressive modelling of noise data is widely used for system identification, surveillance, malfunctioning detection and diagnosis. Several methods are available to estimate an autoregressive model. Usually, the so-called Yule-Walker method is employed. The various estimation methods generally yield comparable parameter estimates. In some special cases however, involving nearly periodic signals, the Yule-Walker approach may lead to incorrect parameter estimates. Burg's method offers the best alternative to Yule-Walker. In this paper a theoretical explanation of this phenomenon is given, while the 1994 IAEA Benchmark test is presented as a practical example of Yule-Walker yielding poor parameter estimates. (author)