SAS for the Macintosh
Welcome to the SAS for the Macintosh Resource Page.
SAS is a large and powerful program for managing and analyzing data; the Mac is a computer that does an amazing amount of stuff. This page is about the intersection of the two.
The goal of this page is to provide information and resources to support users of SAS on the Macintosh. I used SAS on the Mac regularly from mid-1997 until August 1999; I am now working as an application developer in SAS under Windows. I still use both SAS and the Mac OS, it's just that I rarely use them together anymore.
To contact me, email scocca@pobox.com. Let me know if you have any suggestions for additions or changes, some breaking news, or some information you think other Mac SAS users might benefit from seeing.
As part of their Y2K fix rollout, SAS released a minor bug-fix update to SAS for Mac OS. This release identifies itself as 6.12 TS051; SAS has a Fix list on their website. The Mac OS FAQ on their site says that the most recent release is TS070, but as far as I can tell this is incorrect and TS051 is as recent as it gets.
The most frequent question I get from visitors to this page is about the future of SAS under Mac OS. SAS for Mac OS is still only available in version 6.12, although on most other platforms the current release of SAS is version 8.2. SAS has stated that there will be no Mac OS release of Version 8, however they have not yet added the Mac to their official list of platforms for which no future development is being done.
As for Mac OS X, SAS 6.12 for Mac OS is not a Carbon-compliant application. According to SAS, V6.12 will run under the “Classic” environment of Mac OS X. There are no current plans to issue a newer version of SAS for Mac OS.
It is possible that there will someday be another version of SAS under Mac OS. Mac OS has not been added to the official list of platforms for which there is no longer any development taking place. Since there was once a release of SAS for the NeXT platform, some of the basic groundwork for a Mac OS X port may already be in place. In addition, SAS has a strong record of supporting and developing for Unix systems and has recently released SAS for Linux. So while they are not releasing Mac OS versions now, SAS has by no means become a Windows-only vendor.
I am disappointed, but I cannot fault SAS for this decision. They made the effort to produce a Mac OS version in the first place, and the decision to stop making new releases was a financial one based on a relative lack of revenue from Mac OS licenses. This is in part a reflection of the scarcity of Mac OS in the business world; while Mac users in higher education took advantage of SAS, there was very little corporate use of the product. Since the vast majority of SAS revenue comes from corporate licensing, it was not financially reasonable for them to continue with the Mac OS version. I sincerely hope that over the next few years the market will shift so that SAS decides a Mac OS release will be profitable; if that happens, I feel comfortable that SAS will choose to take advantage of the potential revenue stream.
One alternative that still exists, especially if you mostly use SAS occasionally, is to run the Windows version of SAS using an emulator such as Virtual PC from Connectix. I did this for a while (in 1998 and 1999) with Virtual PC version 2 and SAS Version 7 on a Power Mac G3/233 desktop. Performance was not quite good enough to work comfortably with SAS for any length of time, but SAS programs executed reasonably well and I was able to use the emulation well enough to create screen shots for the documentation I was writing at the time.
Both Virtual PC and Apple hardware have increased in speed since that time, while as far as I know SAS has not gotten significantly slower. Thus, it may well be that Virtual PC 4 on G4 hardware provides a good enough user experience to use that as a means of running the current version of SAS under Mac OS. Anyone who has tried the Virtual PC/ SAS combination on a newer Mac is invited to let me know how well it works; I will try to post any results or comments that I get to help you know if this would be an acceptable choice for your situation.
Here are two script packages to make SAS on the Mac easier to run. Click on the icon (or select the substitute link in non-graphical browsers) to download the package described. In some browsers, you may want to option-click to download the files. (These script packages should work properly with SAS 6.12 and with SAS 6.10.)
![[Drop SAS]](dropsasicon.gif){kind=small}
The
Drop SAS script (49K) allows you to run a SAS program in "batch mode" by
dropping the program file onto the script icon. The current version is
1.0.1, which is simply version 1.0 with improved documentation.
![[BBEdit SAS
Scripts]](bbsasicon.gif){kind=small}
William Lisowski generously updated the BBEdit SAS Script
Package (176K) to work with BBEdit 6.0, which has a new
scripting interface which caused the BBEdit 4.0 scripts to fail.
The BBEdit SAS Script Package (74K) allows you to use BBEdit 4.0 from Bare Bones Software as a front end
from SAS. The main script lets you submit a SAS job and view the output
from within BBEdit. Other scripts let you use standard Mac dialog boxes to
select files or folders and then automatically insert the full directory
path into your program.
Each script package is a binhexed self-expanding Stuffit archive (i.e., a .sea.hqx file).
I've written some brief reports on some topics of interest or utility to SAS users on the Macintosh.
The performance notes formerly in this space have been removed, and the project to set up a benchmarking process for Mac SAS to see what affects performance has been put on a back burner. It turns out that the vastly different time requirements for PROC MIXED on two machines were due to a slight variation in the PROC MIXED code rather than to machine configurations. This variation did not affect the estimates produced by SAS, but did increase run time up to twentyfold.
The difference involves the form of the RANDOM statement. The “fast” version used
random x1 x1*x2 ;and the “slow” version used
random intercept x2 / subject=x1 ;where x1 and x2 are class variables. These are the same model; they produce the same estimates. However, the difference in computation time is significant. Somewhat confusingly, SAS documentation suggests that the second method can actually be faster!
It turns out that the key is the number of levels of the class variable. In my test data, x1 had five levels (x2 had about 45 levels, and there were about 2000 observations). It turns out that factoring out x1 (as shown in the second example) is faster only if x1 has a relatively large number of levels. If x1 has a small number of levels, factoring out x1 can, as I saw, greatly increase the execution time.
This is not Mac specific (similar increases show up under Solaris), but it's worth knowing if you use PROC MIXED.
Here are some links to parts of the SAS web site of particular interest to Mac users:
Let me know what you think of this site via email (scocca@pobox.com).
In maintaining these pages, I do not work for or speak on behalf of SAS.
SAS and the SAS System are trademarks of SAS. Macintosh is a trademark of Apple Computer. Other brand and product names used herein are trademarks of their respective companies.
This page was developed on 
Macintosh, using 
Alpha 6.5
and 
Color It! 3.2.
Page last updated 7/21/1.