05-25-2017 06:28 AM
I have build a forecast model, finding that the result of the residual white noise test as below attachment,and all the model's parameters are significant，so what does it mean? does it means there are still some useful information in resudial series?how can i improve the model?
Looking forward to your reply.
05-30-2017 06:03 PM
Indeed, it seem that the residuals has some residual structure (pardon he pun). Given that there is a high peak at lag 12, I am assuming you have monthly data and it has a seasonal component. You could try adding a seasonal factor in your model. Without knowing more about your data and model it is not easy to make more detailed suggestions.
This paper has an overview of how to use the diagnostic plots with a few examples. You can look at the section on PROC AUTOREG for more details.