Unified approach to best unbiased estimation in control problems and linear models

An analog of a sufficient statistic is proposed in the finite-dimensional linear space of statistics. It is shown that the unbiased estimator of the linear function of a parameter closest to a given function, expressed in terms of a minimal linearly sufficient statistic, is potentially unbiased and optimal.Translated from Statisticheskie Metody Otsenivaniya i Proverki Gipotez, pp. 51–58, 1986.