المستخلص: |
A Bayesian method for estimating autoregressive-moving average (ARMA) models is proposed. The proposed methodology is based on replacing lagged errors of the original ARMA model with appropriately lagged residuals from a long auto regression .Bayes estimators have been developed under non-informative and natural conjugate priors. Moreover, the estimation of the parameters for the ARMA representation can be obtained by using a non-Bayesian method, namely the three- stage least squares (3SLS) method. These methods are compared using simulated and real data. A numerical comparison between Bayesian and non-Bayesian will be carried out .It has been seen that the obtained estimators not available in a compact form, although they can be easily evaluated numerically. Moreover, the proposed method (Bayesian procedure) produce estimates with greater precision (smaller MSE) than that for the 3SLS procedure.
|