Statistics and Social Sciences

New Statistical Software

Frank LoPresti

"What groups are you comparing?" I asked. "What is your grouping variable?" She didn't know what I was talking about.

"Perhaps you should read your stats book. . . or meet with your professor." It was Friday afternoon, 2:30. Her assignment was due on Monday and she was not feeling good about her situation. I couldn't spare the time to review all her stats from last year. How could I help her?

I had been reviewing Activ-Stats, a new instructional CD-ROM from a Cornell professor, Paul Velleman. I considered setting her up on it. It would be a first for me, but I try to be flexible. As if we do this with students all the time, I popped Velleman's ActivStats CD into a Pentium in the next office, and she started clicking away.

For the rest of the afternoon, I could hear Velleman's voice cheerfully going on and on. He was tireless, unaffected by her repeatedly asking him to explain which variables in the spreadsheet were categorical, which were quantitative and what that had to do with a t-test. Before I left, she came back into my office. I showed her how to run the test in SPSS. Happy ending, courtesy of Paul Velleman.

I knew Paul Velleman soon after he had studied with Professor John Tukey at Princeton, almost 20 years ago. As a young statistics professor at Cornell's Industrial Relations School, he helped develop modules for Minitab, one of the first interactive statistical packages available. It ran on mainframe IBM computers using paper teletype terminals. He helped write Minitab's stem-leaf and other robust statistical procedures. He went on to create DataDesk, the full-function statistical package which is used in the ActivStats CD-ROM.

ActivStats is an introductory multimedia statistics course. It builds a context for statistical work and prepares the student for the Advanced Placement Statistics Exam. ActivStats uses video, simulation, animation and narration. It is loaded with material from leading textbooks, and has the functionality to allow a teacher to insert her own text and examples. It uses interactive experiments to develop datasets. For example, students use the mouse to "shoot" at a target to develop a dataset, first with the right hand and then with the left. This creates two groups of data, right hand shots versus left hand shots, for the t-test. We compare the average distance from the target's center for each group.

Homework and quizzes follow each chapter. Homework is done within DataDesk, or the student can use the package of the teacher's choosing.

I used to say that, without preparation, I could teach a good statistics course by assigning the Minitab book and telling the students to sit at a computer and follow the book. Now, I'll say that I would assign Velleman's ActivStats CD. Addison-Wesley distributes the CD for about $80, with a discount for students. See their website, www.awi.aw.com, for more information.

Stata has powerful data manipulation and management features such as a full set of numeric and string functions, I/O procedures, appending and merging tools. Since it has full data dictionary functions such as labels and report writing, it is often used as an archive management package.

Stata has a complete matrix language suitable for teaching. It warns that general tools shouldn't be developed because of limitations in dataset size. For advanced programming, the macro language features a large tool chest. This programming language includes the usual commands such as "if" and "while," but advances past SPSS into program debugging, parsing of program arguments and advanced window control to create dialogue boxes and menus. In the usual mode of operation, Stata has windows for output and a spreadsheet-like data editor. Unfortunately, on the PC these windows function differently than they do in most other Windows-based applications. For example, the output window doesn't have a scroll bar to allow the user to scroll the window forward and back. Also, the little "x" in the upper right-hand corner of the data editor doesn't close the window, as it does in most Windows applications. Instead, clicking on it saves the data, exits data editing and brings you to the command prompt, which can be alarming if you simply wanted to close the window. This proprietary window functionality is also a fault of SAS. Although Stata operates in a Windows environment, it also functions by entering commands at the command prompt. It does not have pull-down menus for statistical procedures or for data management.

A version with slightly limited functionality is also available for students. The student version doesn't allow the user to print the graphs produced. Also, it limits datasets to 25 variables and 160 observations. As it runs in DOS, tasks are performed by typing at the prompt. Despite this, the student version is popular, and is available for about $34. Order from their website at www.stata.com.

Stata makes full use of the Internet. The group offers NetCourses -- pay lectures with e-mail responses to questions. Lectures cover Stata techniques and various statistical courses, and involve the participant in a student/teacher dialogue. The cost is about $20 per week. The company's website has many FAQ (Frequently Asked Questions) pages, and the Stata listserv includes useful tutorials.

Stata is the third most often mentioned statistical package, after SAS and SPSS, in circles where such things are discussed. It has an established following. Macintosh, PC and Unix versions are available.

Posted January 20, 1998