Best Linear Predictors in a Stationary Second Order Autoregressive process by means of near and far observations

In this paper, some predictors for prediction in a stationary second order autoregressive process are introduced. The paper attempts to find the best predictor for some cases such as circumstances there exist a fixed number of observations near or far from desired time. Pitman’s measure of closeness and mean square error of prediction are used in order to comparison these predictors. The Gaussian and Gamma distributions have been used for distribution of errors. Finally analysis of two real data sets has also been presented for illustrative purposes.