topic Seasonal trend decomposition by Loess (STL) in SAS Forecasting and Econometrics

Seasonal trend decomposition by Loess (STL)

KZ1 — Mon, 29 Apr 2019 07:15:32 GMT

Hi,
is there any proc statement in SAS that enables to use STL (like in R).

I found a source code for STL function in R but unfortunately I am not able to implement this solution in SAS.

Here are links to the source code
https://svn.r-project.org/R/trunk/src/library/stats/R/stl.R

and the article

https://www.wessa.net/download/stl.pdf

Thank you in advance for your help

Re: Seasonal trend decomposition by Loess (STL)

PaigeMiller — Mon, 29 Apr 2019 10:43:11 GMT

I am not familiar with STL, but of course SAS does have PROC LOESS that might be able to do this.

PROC ARIMA can fit seasonal time series, see https://documentation.sas.com/?docsetId=etsug&docsetVersion=14.2&docsetTarget=etsug_arima_examples02.htm&locale=en

Re: Seasonal trend decomposition by Loess (STL)

rselukar — Tue, 30 Apr 2019 15:37:00 GMT

As far as I can tell, PROC LOESS by itself might not give you the decomposition you want (it could be used recursively to within the overall algorithm). PROC ARIMA also is not directly useful for this. You can use PROC UCM (both ARIMA and UCM are in SAS/ETS) to get such a decomposition. See the getting started example in the UCM doc: https://go.documentation.sas.com/?docsetId=etsug&docsetTarget=etsug_ucm_gettingstarted01.htm&docsetVersion=15.1&locale=en

Also see:

https://support.sas.com/resources/papers/proceedings09/306-2009.pdf for general info on UCM modeling.

http://support.sas.com/resources/papers/proceedings17/SAS0456-2017.pdf for how to adjust for outliers.