![]() | Only 14 pages are availabe for public view |
Abstract The main goal of this study is to solve the identification problem for double seasonal autoregressive models from Bayesian point of view. Two Bayesian identification techniques are employed; namely the direct and the indirect. A 585 simulation studies are conducted to assess the efficiency of both proposed Bayesian techniques. They are also compared with non Bayesian one (Akaike Information Criterion: AIC) taking in consideration the affected factors. These factors include the model order (p, P₁, P₂) the sampling variance (x⁻¹), the series length (n), the seasonal periods (s₁, s₂), and the model coefficients ({u0264}). Results showed that the indirect technique is superior to direct one. Finally the Bayesian identification approaches are applied on six real time series data |