ARIMA degree of differencing

[Dima1]

Hi All

1. Could you please confirm what kind of method Alteryx use to count degree of differencing of data (both seasonal and nonseasonal) when we to not mark any ?

I read that it is KPSS method, but unfortunately I lost that link where i read it and then did not find it.

2. Does Alteryx count degree and than find the best model based on AICc OR Alteryx simple test all possible models (for example 000, 001, 010, 100, 011, 101,110,111 etc.) and then based on AICc find the best?

Thank you for you help!

[CristonS · Accepted solution]

Hi @Dima1

Yes, if the order of first-differencing is missing, it will choose a value based on KPSS test.  If the order of seasonal differencing is missing, it will choose a value based on OCSB test.

You can find more information on the methodology in the documentation for the CRAN forecast package, and auto.arima function.

https://cran.r-project.org/web/packages/forecast/forecast.pdf