Pitfalls in polyspectra

The area of higher-order statistics (HOS) has witnessed a tremendous resurgence over several years. However, Huber, Kleiner, Gasser and Dumermuth's (1971) observations were that 'the newcomer to the field still has to learn the hard way, from his own mistakes. . .' and 'all the pitfalls known from ordinary spectral analysis occur here too, some of them with new twists. . .'. These continue to be true, despite at least two tutorials, and several special issues of journals. The author examines some of these pitfalls, particularly in the context of deterministic signals observed in noise.<<ETX>>