A Direct Bayesian Methodology to Identify the Order of Moving Average Processes using Different Prior Distributions | ||||
| The Egyptian Statistical Journal | ||||
| Article 1, Volume 66, Issue 2, December 2022, Page 1-16 PDF (795.39 K) | ||||
| Document Type: Original Article | ||||
| DOI: 10.21608/esju.2023.164672.1008 | ||||
 View on SCiNiTO | ||||
| Authors | ||||
Mohamed S. Al Bassam; Emad E. A. Soliman   ; Sherif S. Ali
| ||||
| Department of Statistics, King Abdul Aziz university, Kingdom of Saudi Arabia (KSA) | ||||
| Abstract | ||||
| The current study handles the direct Bayesian identification of moving average processes based on different priors. The priors considered in the article are g, natural-conjugate as well as Jeffreys' priors. Posterior probability mass functions for moving average model's order are developed using the above mentioned three priors. Then, the order with maximum probability is selected as the identified model order. In order to ease computations, the likelihood function of the moving average process is approximated. Effectiveness of each posterior mass function in identifying an order for the model is assessed via Mont-Carlo simulations. Numerical results support the use of the proposed Bayesian technique to solve the considered identification problem. Furthermore, the performance of the technique using g prior is better than its performance using the natural-conjugate prior which, in turn, is slightly better than the performance using Jeffreys' prior | ||||
| Keywords | ||||
| Moving average processes; Direct Bayesian identification; Indirect Bayesian identification; Informative priors; Non-informative priors | ||||
|
Statistics Article View: 214 PDF Download: 409 |
||||
