Indirect and direct bayesian techniques to identify the orders of vector arma processes | ||||
The Egyptian Statistical Journal | ||||
Article 2, Volume 62, Issue 1, June 2018, Page 15-34 PDF (6.06 MB) | ||||
Document Type: Original Article | ||||
DOI: 10.21608/esju.2018.244222 | ||||
View on SCiNiTO | ||||
Abstract | ||||
This article develops two bayesian techniques to identify the orders of vector mixed autogressive moving average processes namely the indirect and direct techniques. The proposed indirect technique approximates the joint posterior probability density function of the coefficients of the largest possible model by a matrix t distribution. Then by employing a series of tests of significance the insignificant coefficients are eliminated and the model orders are determined. On the other hand the proposed direct technique derives an approximate joint posterior probability. A wide simulation study is conducted to examine the effectiveness of the proposed procedures and compare their performance with the well-know ALC technique. The numerical results show that the proposed techniques can efficiently identify the orders of vector autoregressive moving average processes for moderate and large time series lengths. Moreover the indirect technique dominates the direct and ALC ones | ||||
Keywords | ||||
Vector ARMA processes; indirect bayesian identification-direct bayesian identification- posterior probability mass function; matrix normal; wishart prior- Jeffreys prior | ||||
Statistics Article View: 53 PDF Download: 134 |
||||