topic Re: PROC REG Forecasting (Creating Forecast Confidence Intervals) in SAS Forecasting and Econometrics

PROC REG Forecasting (Creating Forecast Confidence Intervals)

Kevin_L_ — Sat, 09 Feb 2019 17:57:23 GMT

I want to create a AR(P) and eventually ARDL forecasting models by estimating the parametar values via OLS. SAS seems to only use MLE or other non linear least squares to fit models. PROC AUTOREG seems to use OLS up until AR terms are used in the model (OLS isnt an option with nlag=). Therefore, I am using PROC REG and haveing to create forecasts myself. Most of that is easy except for the h-step-ahead forecasting intervals. Anyone know of a good way to program those? Or is there a package that estimates via OLS and will forecast?

Thanks.

Re: PROC REG Forecasting (Creating Forecast Confidence Intervals)

imvash — Sat, 09 Feb 2019 22:01:13 GMT

I suggest using PROC ARIMA. See the documentation at:

http://support.sas.com/documentation/cdl/en/etsug/63939/HTML/default/viewer.htm#etsug_arima_sect001.htm

You can specify LEAD=n option in FORECAST statement. See the details at:

http://support.sas.com/documentation/cdl/en/etsug/63939/HTML/default/viewer.htm#etsug_arima_sect024.htm

See the estimation methods at the following link; you might want to use the METHOD=ULS option as it produces unconditional least squares estimates:

http://support.sas.com/documentation/cdl/en/etsug/63939/HTML/default/viewer.htm#etsug_arima_sect036.htm

Alternatively, you can use the combination of HPFARIMASPEC procedure and PROC HPFENGINE. You will have all the options specified in PROC ARIMA as well. See the details at:

https://documentation.sas.com/?docsetId=hpfug&docsetVersion=14.3&docsetTarget=hpfug_hpfarimaspec_overview.htm&locale=en

Re: PROC REG Forecasting (Creating Forecast Confidence Intervals)

Kevin_L_ — Mon, 11 Feb 2019 13:45:58 GMT

Thanks, Imvash. Is ULS the same as OLS? From the documentation, it appears to be same. Or is there a difference?

Thanks.

Re: PROC REG Forecasting (Creating Forecast Confidence Intervals)

imvash — Mon, 11 Feb 2019 13:50:48 GMT

Same. It minimizes the sum of squared residuals.

Re: PROC REG Forecasting (Creating Forecast Confidence Intervals)

Kevin_L_ — Mon, 11 Feb 2019 15:53:11 GMT

Thanks again. One more question: When I run the same model in proc reg and proc arima, why would the parametar estimates differ? They are only differenent if I include lagged and differenced dependent variables. Does a data step calculate the lags and differences differently than proc arima? Or what would be the difference in estimates? Thanks!

Re: PROC REG Forecasting (Creating Forecast Confidence Intervals)

kamalhossain22 — Tue, 30 Nov 2021 14:29:03 GMT

Dear sir, good evening. I am new to SAS econometrics. I want to run Autoregressive Distributed Lag (ARDL) Model in SAS, but I could not find the steps to run the ARDL model in SAS. I could not find any example of the ARDL model in SAS, Can please help me provide the steps to run ARDL in SAS? Thank you with regards.