Using the table above we can run the correlelogram function on Eviews to obtain the following picture:
We can see from the Correlogram of the first difference that this data generating process for the series has a few possible candidates for models.
ARIMA(1,1,0)
ARIMA(1,2),1,0)
ARIMA((1,4,2),1,0)
ARIMA(1,4),1,0)
We will estimate these models and use the AIC and SBC statistics in addition to t-statistics to choose the model which most closely approximates the data generating function. We estimate the largest one and use t statistics to eliminate the irrelevant variables
ARIMA((1,4,2),1,0)
ARIMA(1,4),1,0)
ARIMA(1,2),1,0)
ARIMA(1,1,0)
The model which had the highest adjusted R-square where all the regressor’s estimated parameters where statistically significant is the ARIMA(1,4),1,0). This model is also the one with the lowest AIC score from the models whose regressors are statistically significant.
Share with your friends: |