Determining the order of differencing and the true generating model using information criteria for Arima (0,1,1) and Ar (1) processes

Abstract

When a unit root test is applied to an ARIMA (0, 1,1) process with large and moderate values of the moving average parameter serious size distortions will appear. Contrary, the test has almost no power if it is implemented to an AR(1) process with large and moderate values of the autoregressive parameter. This study investigates the possibility of determining the order of differencing and the true generating model for these two simple processes using the AIC and the SBC information criteria and it finds that the use of these criteria will assist the analyst to successfully determine the order of differencing and the true model.