In this investigation the geneal solution is derived for the problem of the optimum linear estimation of a sampled stochastic process, when the transition and output matrices of the model of the process are random parameters that are independent from one sample point to the next with known mean and covariance. The resulting estimate is optimum in the sense that it minimizes the trace of the covariance matrix of the error (a generalized mean-squared-error criterion). All of these results are derived from the sampled version of the Wiener-Hopf equation, and they apply without modification to stationary and nonstationary statistics and to growing-memory and infinite-memory filters.