Stationary reversibles processes MA and ARMA

Abstract

In this paper, we show how to choose an adequate model of a reversible stationary moving average process of finite order, given the appropriate number of sample correlations. For an order no higher than the fifth, a one-to-one correspondence between the coefficients and correlations of the process is established. For higher-order moving average processes, a mixed autoregression and moving average model is preliminarily fitted to the initial data. This option also has an independent meaning.