Parameter Estimation In The Presence Of Low Rank Noise
R.T. Behrens, Louis L. Scharf · 2005
We extend the classical problem of parameter estimation in the linear model to the case where low rank, or structured, noise is present in addition to the usual full rank noise. Many of the solutions are written in terms of oblique projections, a natural extension of the orthogonal projections of the classical case. The obliqueness of the pro jections stems directly from the non-orthogonality between the signal subspace and the structured noise subspace. We show how rank reduction techniques may be applied simul taneously to the signal and structured noise subspaces.