A Multi-Stage Procedure of Bayesian Estimation for First Order Moving Average Model
Abstract
In this paper we present a multi-stage procedure for improving the Bayes estimates of first order moving average model (MA(1)) using the estimated residuals. Simulation results based on different model structures with varying numbers of observations are used to investigate the performance of the proposed procedure. The results show a remarkable improvement of the t-approximation using two and three stage procedure in terms of the posterior mean and variance for different model structures. It is also noted that these estimates converge as the series lengths increase.
About this article
References
S. Shaarawy, L. Bromeling, Bayesian inferences and forecasts with moving average processes, Communication in Statistics: Theory and Methods 13(15) 91984) 1871-1888.
M.M. Smadi, Bayesian inference of threshold autoregressive moving average models using sample-based methods. PH. D. Dissertation, Colorado State University, Fort Collins, U.S.A. 1997.
M. Smadi, A. Abu-Taleb, On Bayes estimation of first order moving average model, Journal of Mathematics and Statistics 2(3) (2006) 422-426.
S. Koreisha, T. Pukkila, 1990, A generalized least-squares approach for estimation of autoregressive moving-average models, Journal of time series, 11(2) (1990) 139-151.
T. Safadi, P.A. Morettin, Bayesian analysis of threshold autoregressive moving average models, Sankhya., 62(3) (2000) 353-371.
View more references (2)
W.S. Cathy, Bayesian inferences and forecasting in bilinear time series models, Communication in Statistics: Theory and Methods, 21(6) (1992) 1725-1743.
M.M. Smadi, M.T. Alodat, Bayesian threshold moving average models, 10(1) 2011 (262-267).