Linear dynamic errors-in-variables models. Some structure theory (Q911205)

The paper surveys recent results on identification problems for linear dynamic errors-in-variables-models. First, the authors state the general problem under the assumption that the noise components of the model are mutually uncorrelated. This section is followed by the analysis of single-input-single-output-systems. After formulating the model, the authors present causal solutions and conditions for identifiability. Part 4 of the paper is concerned with the three-variables case and part five extends this type of model to vector-input-vector-output-systems under the assumption that input and output noise are uncorrelated. In the final section emphasis is given to the non-Gaussian case and it is shown how higher-order moments can be used for identifiability.