Unicast

Right now I'm taking an introductory course on theoretical statistics. Most of this semester's course is focused on teaching the ubiquitous statistics package SAS. This will be quite useful because, at least in Denmark, most companies that do any kind of statistics use SAS.

SAS is ridiculously expensive. A basic license for SAS Foundation can easily cost $5,000/person/year. Many Danish companies and government agencies pay more than 100% of their yearly payroll costs for their data analysis staff to SAS, every year. Even though software like SPSS and S-Plus, whose feature sets are comparable to the statistics portions of SAS, are cheaper, people are still looking for alternatives.

The latest edition of R News (664k PDF) has, starting at page 2, an account by the president of MedAnalytics, Marc Schwartz, on why he chose R — an open source version of the venerable S statistical system — for his new company.

Another good open source program is gretl, the GNU Regression, Econometrics and Timeseries Library, a relatively easy to use tool to browse, manipulate and analyze data sets. It allows you to easily load data, run regressions and tests, perform simulations, plot graphs with gnuplot and create seasonally adjusted timeseries with the bundled version of X-12-ARIMA, the industry-standard seasonal model. Below are graphs of ARIMA output for completed buildings in Denmark from February 1999 to June 2004, which took less than 5 minutes to create in gretl and cost $0 (from the top: seasonally adjusted figure, trend/cycle and irregular component).