Behaviour of the parameter of Non-Guassian Seasonal Autogressive Model from first order SAR (1)

Abstract

This research aims at following the behaviour of Parameter of Non – Gaussion Seasonal Autoregressive Model from first order by simulation. This is done by supposing that model random errors follow Non – Gaussion continuous distribution and by using two comparison tools mean square error MSE ( Φ ) and mean absolute percentage MAPE ( Φ ) for the difference cases (seasonal period , sample size , parameter initial values and the sign of parameter initial values ).
The main conclusions are (i) MAPE ( Φ ) and MSE ( Φ ) values increase when seasonal period is increase , sample size with negative initial value of the parameter and the parameter initial values are approaching zero, and ( ii ) MAPE and MSE values of parameter Φ decrease when season period is decreased (Regardless of the parameter is positive or negative), parameter initial value is ± 1 and sample size is increase with positive initial value of the parameter Φ.