SAS vs S-Plus

From: "Bonate, Peter" Subject: [NMusers] SAS vs S-Plus Date: Wed, December 8, 2004 2:35 pm I guess it's time to stir up the pot a bit. While I agree with Mike Fossler that every modeler should learn a higher-order programming language (and there are many: S-Plus, IML in SAS, Matlab, Gauss, O-matrix), I disagree that S-plus should be the language of choice. Matlab, Gauss, and O-matrix are useful but they all suffer in that statistical analyses are not easily implemented. In the case of Matlab and O-matrix additional modules must be purchased to do any kind of statistical analysis so they aren't really useful for what we do. Also, they don't do mixed effects modeling. So really it comes down to SAS and S-plus. S-plus early on became the program of choice because it had better graphics, it still does over SAS, and has some procedures like GAM that SAS did not. However, the analysis gap has closed - both offer nonlinear mixed effects models, GAM analysis, etc. They are virtually the same. S-plus still has the edge graphics wise. However, of all the programming languages I have studied, S-plus is one of the craziest and not the most easy to learn. SAS is far easier to learn and has the advantage in the all statistical analyses in support of an NDA will be done within SAS. Most companies have a site license for SAS and so using it will not cost a department much, if anything. S-plus is a couple of thousand dollars. I also have to say that SAS has the regulatory edge in terms of acceptance over S-plus. Comments? Pete Peter L. Bonate, PhD, FCP Director, Pharmacokinetics ILEX Oncology 4545 Horizon Hill Blvd San Antonio, TX 78229 phone: 210-949-8662 fax: 210-949-8219 email: pbonate@ilexonc.com