Automatic seasonal auto regressive moving average models and unit root test detection

Abstract

It is well known that in reality, sequential data more likely exhibit a non-stationary time series or a seasonal non-stationary time series than the stationary one. Therefore, a hypothesis is needed for testing those properties in the time series. Various tests are available in the literature. However in this study, we applied unit root test of Dickey Fuller, Augmented Dickey Fuller and Seasonal Dickey Fuller. We also designed a forecasting program using R 2.6.1. This program executes raw data and gives information of the best time series model in the sense of minimum AIC (Akaike Information Criterion).Using this program, a user who doesn’t have a grounded background in time series analysis will be able to forecast short-period future values time series data accurately. The analysis of data consists of Box-Cox transformations, unit root test, removing unit root and seasonal components, finding the best time series model for the data, parameter estimation, models diagnostic checking, and forecasting of the future value time series.