Econometric modeling: your questions answered


Several weeks ago, I led a SAS Talks webinar on SAS/ETS emphasizing the many recent changes to the software. SAS/ETS, for those unfamiliar with the product, is SAS’s suite of econometrics, time series and forecasting tools and algorithms. While we covered a substantial amount of material in the talk, there is even more that I didn’t have time to share.

The first place to turn for information on the many recent enhancements to the product is the product documentation:

Those resources are great but may I suggest actually putting your hands on the software? As part of the new SAS Analytics U initiative, we have made our cloud version of SAS freely available to those wishing to use the tool for a learning or teaching purpose.

  • If you would like to access this tool, you can sign up for a free account. The next step is to actually submit code for yourself.
  • For those looking to get started with one of SAS’s newest procedures, try these PROC HPCDM code samples.
  • The SAS/ETS 13.2 User's Guide Sample library has fully self-contained examples of how to run all SAS/ETS procedures. Just copy and paste the information on the page into the SAS Studio browser and you are running SAS code in the cloud. Pretty cool!

There were several excellent questions which were asked during my SAS Talks webinar.  Here are some of the questions and responses.

Q: ­If one listens to the hype one might conclude that with "Big Data" tools, time series and explanatory econometrics are no longer needed. Please comment.

A: This is a great question and one that probably deserves a full post. If you would like to hear my take on the importance of econometric thinking and the dangers of simply worrying about prediction, feel free to watch Econometrics--The Question of "Why?".

Q: Are multiple frequency time series equivalent to multiple (multivariate) time series?

A: A time series with multiple frequencies are different from multivariate time series arising from the same system however the storage of these time series might make them look similar. In both cases, to store these values in one data set, additional columns must be created. In the case of multivariate time series it would not be anticipated that the resulting data set have missing values however, in the case of multiple frequency data, it would be expected.

Q: How do we know what version of SAS/ETS we have?

A: An easy way to do this is to look at the log file immediately after your SAS session begins. The versions are listed. An alternative way is to run this code.

proc product_status;

It will list all your currently licensed products and their corresponding versions.

Q: Can you share the code to estimate a count data regression model with household specific fixed effects from the doctor visit example?

A. Sure.  Notice that the only difference in syntax is the groupid= option on the proc statement.

proc countreg data=a groupid=panelid;
   model visits = x1 x2 x3 / dist= poisson;

Thanks so much and I look forward to sharing highlights from future releases with you.



About Author

Kenneth Sanford

Sr Research Statistician

Kenneth Sanford, Econometric Evangelist, Advanced Analytics R&D, SAS: Ken joined SAS after working for both a large private consulting firm and holding faculty positions at the University of Cincinnati and Middle Tennessee State University . At SAS, Ken is responsible for helping to integrate the advanced analytical procedures being created by the ETS (Econometric and Time Series) development team into the business solutions offered by SAS. He is also charged with promoting the use of the newest Econometric estimation techniques to our new and existing SAS users. His background is in applied economics with special interest in Industrial Organization and Health Economics. He has articles forthcoming in the Southern Economic Journal and the Journal of Sports Economics. Ken holds a Ph.D. in Economics from the University of Kentucky.


Leave A Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Back to Top