Tools of the Stats Trade, Part 2

This is the second in a series of articles written by the ACCC's consultant for statistical software, Violeta Carrión, STATS@uic.edu, who expresses her thanks to Dick Campbell who provided the essence of the information about Stata.

Yesssss, the tale of many Ss continues! If you recall, we discussed SPSS and SAS in our previous article. We noted that a main difference between the two is that SPSS offers a more user-friendly graphical interface, good for beginning users, and SAS caters more to the programmer in you. Underlying both packages there is a powerful programming language. (Breathe deeply! ... Yes, if you use SPSS you have been programming all along!!)

You might be thinking, ssssso many packages! Which one should I use when? Should I explore others programs or stick to what I know? Excellent questions! A salient element is the amount of programming flexibility they allow the user. That is, your ability to modify how the program actually works.

And it is on this note is that we start our tour today of two more statistical packages: SYSTAT and Stata. These packages offer the convenience of a graphical interface, but they also emphasize the programming component. They are more challenging to learn, but they also can be much more rewarding.

The child of Dr. Leland Wilkinson, statistician and former professor at UIC, SYSTAT had its roots right here at UIC some years ago. It was in the DOS prompt era; at the time, powerful stats packages were primarily available on mainframe computers. SYSTAT was developed for programmers who wanted to do statistics on their desktop computers; its emphasis is on sound statistical algorithms and cutting edge graphics.

In 1996, SYSTAT was purchased by SPSS, Inc. As a result, SYSTAT's user interface improved significantly. SYSTAT graphics remained at the cutting edge. Earlier this year, SYSTAT Software Inc (SSI), purchased SYSTAT from SPSS. The SYSTAT Web home page is at: http://www.systat.com/

So, what does SYSTAT offer you? SYSTAT offers both a programmer's and a beginner's interface. Upon opening SYSTAT, you are presented with its original interface tailored for command prompt users. In this view, you will find three panes: output, results, and command prompt. Upon opening a dataset, a window with pull-down menus much like those of SPSS opens up. You can use SYSTAT via its pull-down menus and not have to much actual programming.

SYSTAT can perform comparable statistical procedures to other packages, including missing value and power analysis. In addition to other commonly used formats, SYSTAT can open and save files both as SAS datasets and SAS transport files. It can also open ArcView and BMPD files, and save BMPD files. If you want plain text output, you can do it by enabling SYSTAT Classic in the Options menu.

The software's user manuals, especially Getting Started, should suffice to get introduced to using the package. Another good resource is the book Desktop Data Analysis with SYSTAT, by Wilkinson, Blank, and Gruber, 1996.

SYSTAT features:

http://www.systat.com/products/Systat/

Learning SYSTAT:

http://www.marymt.edu/~madden/stsyswn9.htm
http://www.psych.upenn.edu/systat/smallfiles.htm

Stata emphasizes the programming aspect of statistical computing. It can be used interactively, via its command line, or in batch mode. It was designed for economists (and as such, it compares to LIMDEP), for epidemiologists, and for survey-data and experimental analysts. For survey analysis, for example, Stata has routines for complex multistage survey samples and can evaluate patterns of participation in survey data and capture them in a variable. Stata does everything that its competitor SUDDAN does, except analysis of poststratified data.

One of Stata's strong points is its memory management capabilities for computationally demanding routines. It takes the dataset into memory, which allows for faster data access; default memory size can be increased as desired. Like other statistical programs, default parameters can be also modified via a profile file.

When you open Stata, you encounter four windows: results, command, review, and variables. The results window interactively displays information as you submit commands or queries. The command window is to submit syntax, the variable window displays the available variables on the data file you are using, and the review window lists a history of submitted commands. Stata command files are text files called DO files, and when compiled as a program, they are called ADO files.

From experience, I can tell you that it is not hard to learn to use this package after you understand how it behaves and become familiar with its command language. Software updates are easily accessed via the update query command that connects to the Stata Web site and automatically downloads any updates and program files. Pretty cool!

Finally, Stata offers a set of courses via the Web. Information on these courses is available at: http://www.stata.com/info/products/netcourse/

Stata features:

http://www.stata.com/info/whystata/

Learning Stata:

http://www.stata.com/info/capabilities/session/
http://www.ats.ucla.edu/stat/stata/modules/
http://www.princeton.edu/~erp/stata/main.html
http://www.umass.edu/acco/statistics/handout/stata7.pdf

Well, I hope you are sssssatisfied. I will leave you with a word of advice. If you are curious about how to use a program, "Google it!" (http://www.google.com) I frequently find that there is a Web page with the answer to whatever statistics or software usage questions that I have. Many sites include detailed information on how to use a particular package.

Statistically yours,
Violeta Carrión

Stats Support Email Address:

STATS@uic.edu

Stats Support URL:

http://www.accc.uic.edu/stats/

ACCC Software Sales:

http://www.accc.uic.edu/stats/software.html

UICSTATS-L archives:

http://listserv.uic.edu/archives/uicstats-l.html