Under normality, the Bayesian estimation problem, the best linear unbiased estimation problem, and the restricted least-squares problem are all equivalent. As a result we need not compute pseudo-inverses and other complicated functions, which will be impossible for large sparse systems. Instead,...