my candidate model is arima(5,1,5) the output in stata is the picture below. It seems like the other coefficient is not significant. I test the residuals and it is white noise, thus by box and jenkins the model is appropriate. But the other coefficients there is not significant. IS THE ARIMA(5,1,5) IS STILL MY FINAL MODEL?
Why don't you consider other model(s) and compare mean squares? You need better forecasting ability. So just consider other models as well ans compare mean squares. Model with less mean squares will be the appropriate one.
According to the STATA output the proposed model I would recommend to try another model does not seem to be very good (look at the p-values of AR(5) and the intercept, for instance). As Ashraful Khan suggested, you should try another models (including stational ARIMA) and compare their sum of squared residuals, log of likelihood and also checking the residuals .....
- Badges
- Science topic
- Similar topics
- Statistics
- Time Series