Hi,
I think minic can produce the desired info in identify statement of proc arima.
Thanks,
Just some ideas, if you have for example monthly data:
- proc arima and a "Var=<Variable> MiniC P=(0:12) Q=(0:12);"-statement, but you probably never would have a MA(5,7,11) or so and you would have to repeat this for the cases of differencing, seasonal differencing, taking logs, etc.
- de-seasonalize the data using proc expand or proc x11 and use minic of a low order afterwards
- pre-define simple (to avoid over-fitting), but "propable" models (AR(1), AR(1,2), AR(1)(12), etc.) and find the model with the lowest SBC "manually" (or with clever code)
Hi,
in the documentation minic identifies only p and d, it does not identify P and Q, can you please illustrate how to identify P and Q
Hi,
I also consulted documentation and found nothing about P and Q. Sorry about that.
Just some ideas, if you have for example monthly data:
- proc arima and a "Var=<Variable> MiniC P=(0:12) Q=(0:12);"-statement, but you probably never would have a MA(5,7,11) or so and you would have to repeat this for the cases of differencing, seasonal differencing, taking logs, etc.
- de-seasonalize the data using proc expand or proc x11 and use minic of a low order afterwards
- pre-define simple (to avoid over-fitting), but "propable" models (AR(1), AR(1,2), AR(1)(12), etc.) and find the model with the lowest SBC "manually" (or with clever code)
Thank You. This is very helpful.