The optimum value of is given by

           The corresponding minimum linear mean-square error (LMMSE) is

           The orthogonality principle can be applied to obtain optimal estimation of a random variable from more than one observation. We illustrate this in the following example.