On a Decomposition of the Singular Gauss-Markov Model

Abstract

It is well-known that singularity of the dispersion matrix in a Gauss-Markov model may have various consequences, which remain obscure in models furnished with a regular dispersion matrix (see e.g. Rao [9] or Zyskind [16]). Roughly speaking, these consequences appear in form of inherent restrictions on the vector of observations or parameters (or both).