Estimation of the Minimum Mean Square Error of Prediction
最小预测均方误差的估计
Bloomfield (1973) and Jones (1964) have discussed the estimation of the error of prediction of a time series. Their results use the asymptotic normality of their estimates and we attempt to examine the validity of this approximation in the simplest case. We show that one can approximate the distribution of the mean square error in small-sample situations, and in consequence obtain some idea of the rate of convergence to normality.