“…Second, we can prove α kk and β kk are best quadratic unbiased estimator (BQUE) for α kk and β kk for every k and k . We need to check the conditions of Theorem 2 in Zmyślony (1976), or the lines of analogous arguments in Roy et al (2016) and Koziol et al (2017), that is, given P 0 V = V P 0 and R 0 = I pn − P 0 , there exist BQUE for the parameters of quadratic covariance if and only if span (R 0 V R 0 ), i.e., the smallest linear space containing R 0 V R 0 , is a quadratic space. Since Σ(A, B, τ ) is a UB matrix, {Σ(A, B, τ )} 2 is a UB matrix, expressed by Σ 2 (A 2 , AB + BA + BP B, τ ), so span{Σ(A, B, τ )} is a quadratic subspace and the identity matrix I(τ ) belongs to it.…”