Stationarity in an Interrupted time series

Levi_M — Wed, 03 May 2023 15:33:22 GMT

I am using proc autoreg in SAS to conduct an ITS analysis and I have a question about stationarity. Proc autoreg is able to perform the augmented dickey-fuller (ADF), the phillips-person (PP), and the KPSS test for stationarity. I believe I have a good understanding of the difference in their interpretations.

My question is whether I test for stationarity across the interruption. For example, if I had a 36 month study period with an interruption at month 25, would I look at stationarity as separate parts (months 1-24 and 25-36) or as a whole (1-36)? My thought process is, that the interruption would likely cause it not to be stationary and thus ran separately.

Thank you for the clarification

Code I am using below, but this seemed more of a statistical approach logic rather than a code question:

proc autoreg data=one;
    model outcome = /stationarity=(kpss=(kernel=qs auto))stationarity=(phillips);
where  1<= month <= 24;
run;

Re: Stationarity in an Interrupted time series

sbxkoenk — Thu, 04 May 2023 15:12:39 GMT

Hello,

I have moved your question to the "SAS Forecasting and Econometrics" - board.

Stationarity is never an end in itself (an objective in itself).

The interruption / structural break in your time series will indeed probably break the -- possibly present -- stationarity as well (since "stationarity" means that the statistical properties of a time series -- or rather the process generating it -- do not change over time).

But the question is : do you need to address non-stationarity?
OK , stationarity is important because many useful analytical tools and statistical tests and models rely on it. But what is your end-goal? What do want to achieve?

Koen

topic Re: Stationarity in an Interrupted time series in SAS Forecasting and Econometrics

Stationarity in an Interrupted time series

Re: Stationarity in an Interrupted time series