View article

Consider a model AX= B, where A∈ Rm× n, X∈ Rn× p, and B∈ Rm× p. The model means the following. For the true values, we have AX= B, where A, B, X are non-random matrices. We observe A= A+ A and B= B+ B, where A, B are random matrices. We want to estimate X with fixed n and p and increasing m. This general model includes static models if the rows of [A, B] are independent, and dynamic models if the matrices [A, B] and [A, B] are structured. Common assumption is that the covariance structure of [A, B] is known up to a scalar factor, which leads to Elementwise–Weighted TLS Problem in the static case and to Structured TLS Problem in the dynamic case, see Kukush and Van Huffel (2004) and Kukush et al.(2005b). In the present paper we assume that [A, B] is partitioned into two uncorrelated blocks [D1, D2] of sizes m× n1 and m× n2, and EDT k Dk= λ0 k