I was employed with two different companies that used SAS in support of federal contracts and did consulting work for another. One was with DOD, the others were with DOE, DHHS, and Agriculture.

In every case, the contracting companies had an extension of the agencies' licenses written into the contract to cover machines to be used off-site.

Admittedly, this was a long time ago (1990-1996 or so). It involved a lot of paperwork.

In any case, with licensing charges that steep, you need to build it into your rates. I don't know what your rates or pipeline looks like, but if you're expecting 1000 billable hours per year, that's six more bucks an hour (about four if you take your tax writeoff into account--but it's not a good idea to mix pre- and post-tax in your accounting, though).

As for what I'd be willing to pay for professional tools, my answer would be "very little." There is so much good stuff available for free nowadays (with open source and commercial development-only licenses) that in any case where it is necessary to purchase a license over about $500 or so to do work, I would try to pass that charge through on the contract.

Of course, I have no idea how I would handle the situation if one package was the core of my business.

I think I have a better grasp now: You would be using SAS to analyze data for clients.


When I read your original post, I was under the impression that you would be providing SAS programming services for clients who use SAS. Correct me if I'm wrong, but this makes you an "end-user" of SAS. (?)

My recommendation would be to do an honest analysis on what the impact of purchasing a license would be on your rates and whether your market will accept those rates.

I don't know what SAS tools you intend to use specifically, but a quick, nosy peek at your resume (please forgive the intrusion) tells me that you could cobble-together Perl code to handle most of the BASE and STAT functionality and find acceptable libraries to handle your GRAPH and SCL needs (I'm assuming that SCL is Screen-Control Language--my memory is hazy here). It wouldn't be the three-pane IDE, but a savings of $6k might do a lot to help you not miss it so much :-)

Are you aware of the Statistics:: modules?

[link|http://search.cpan.org/search?query=statistics|http://search.cpan.o...?query=statistics]

GNUPlot is rough, but I've used it in the past for generating IVR call statistics graphs. Also, OpenOffice.org has a graphing component in its spreadsheet--I haven't used it, though.

One thing that was weak in SAS when last I used it was its interface to SQL back-ends. IMHO, DBI is awesome. I think that quite a bit of DATA step and MERGE pokiness could be overcome with a SQL back-end.

After all this babbling, I guess my bottom line is that this may be an opportunity to see if a couple thousand dollars' worth of hours thrown at developing yourself to the point of obviating SAS would make sense for the long haul.

First, sorry to have missed the gestalt references.

Second, I've now skimmed your manifesto.

I dunno that I can contribute anything of use to you since you picked Python and MySQL while I'm strictly Perl and PostgreSQL.

As for the "rebuild SAS" scenario, I'm not certain that SAS is the right approach to data analysis. The few places I've used/supported it, there were always a few FORTRAN guys on staff to handle tricky stuff that could not be done with SAS. I also remember that there was always a go-to guy to do data cleanup. Heck, when I was working at a DHHS agency, I developed several tools (under scientist supervision) with TurboPascal to avoid enormous CPU-usage costs at the NIH datacenter.

I know that SAS is the Cadillac standard, but it might make sense for you to pull-together the Python and MySQL tools and perform data analysis "by-hand" to for some of your work just to see if it can be done. If you tWikify all your steps and code, you may find what steps are worth "automating" and maybe get some review of your methods at the same time.

Again, my take would be to consider how much time you could afford in lieu of paying an annual recurring $6k.