SYSTAT was first developed in the late 1970s by Leland Wilkinson, then a statistics professor at the University of Illinois at Chicago. The first PC/DOS version was produced in 1984, just three years following the sale of the first IBM/PC in August 1981. Wilkinson’s goal was to offer researchers a PC/DOS environment in which to develop and estimate a suite of descriptive statistics as well as a number of the most well-used statistical models. At the time there were few DOS-based statistical packages; major packages such as SAS and SPSS were still mainframe oriented.

Wilkinson fashioned theSYSTATlanguage and program style after that of SAS. Even some quarter of a century later SYSTAT still has a SAS flavor, particularly its programming language.

In 1987 SYSTAT introduced a separate graphics package called SYGRAPH. It was perhaps the most sophisticated statistical graphics package available at the time. Even though it was DOS based, its graphical capabilities far exceeded those produced using SAS and SPSS, who by then had come out with their own PC software. I first started to use SYSTAT in the late 1980s largely because of its associated graphics package.

The Windows version of SYSTAT, called Version 6 for Windows, did not appear until 1996, a bit late in the competitive market of the period. I believe that its slowness in entering the Windows environment resulted in a substantial decrease in user base. Other packages such as StatGraphics, Statistica, Stata, and even SPSS had earlier exploitedWindow’s graphical-user interface (GUI), and had developed sophisticated graphical capabilities that were unavailable in the DOS environment. Even Macintosh-based packages such as Data Desk provided the user with an interactive graphical user interface.

Over the years SYSTAT has added a host of new statistical capabilities. Unfortunately, however, SYSTATs advancements have not been as well recognized as have the enhancements made by other statistical packages. Partially this has been due to its late entry into theWindows environment—once a package falls behind it is difficult to resurrect. But other reasons have occurred as well. SYSTAT changed ownership several times during the 1990s and never received the appropriate marketing push to keep its name in the forefront. Perhaps the effortwas made, especially with Version 6 for Windows, but many researchers had by then moved on to other packages. I must admit that I was one who ventured off to other software.

Again, SYSTAT has continually been improving its statistical offerings as well as its interface. Over the years it has retained a loyal core of users who would not think of needing another package for their statistical work. With the 2004 release of Version 11, however, I believe that SYSTAT can and will attract a host of new users. Its list of capabilities is impressive, and the ease with which one can convert files saved in other major statistical packages into SYSTAT format makes it one of the more user-friendly packages on the current market.

• Windows® XP; 2000, NT; ME; 98
• Pentium® level processor
• 64 MB RAM
• 150 MB free disk space, plus 80 MB manuals pdf
• CD-ROM; SVGA adapter & monitor

The student version consists of SYSTAT version 10.2, the last upgrade. SYSTAT has made the decision to provide the fullcapacity software to students. There are none of the limitations of reduced observations and variables, or of the printing and storage of results that is typical with student versions. Students have full access to the manuals as well, in pdf file format. I know of no other major statistical software company that is so student-friendly.


2. VERSION 11 CAPABILITIES

Because SYSTAT has not been reviewed in The American Statistician for many years, I believe it worthwhile to list some of its major capabilities.

2.1 Descriptive Statistics

SYSTAT has the full repertoire of standard descriptive statistics. Additionally it allows calculation of N&P tiles, resampling statistics including bootstrap and jackknife, column and rowcalculation of first-order statistics, fitting of 7 discrete distributions, 18 continuous distributions, and QuickGraph graphs of respective observed and expected frequencies.


2.2 Tables and Regressions

Multiway tables with all standard row and column statistics; correlations, distance, and similarity statistics; canonical correlations; missing value analysis; loglinear models; regressions including OLS, Bayesian, ridge, robust, logistic, probit, nonlinear, two-stage LS, and mixed (hierarchical linear models with fixed or random effects, autocorrelated errors structures, twolevel nested model analysis, and other features).


2.3 Other Statistical and Graphic Capabilities

SYSTAT allows the user 126 nonparametric smoothers, including LOESS; hypothesis testing; ANOVA; general linear models; quality analysis; nonparametric tests; MANOVA; factor and principle components analysis; discriminant analysis; cluster analysis (hierarchical, K-means, additive trees); correspondence analysis; classification and regression trees; test item analysis; multidimensional scaling; partially ordered Scalogram analysis with coordinates; perceptual mapping; conjoint analysis; time series analysis; spatial statistics (2D/3D variograms, kriging and simulation, etc.); signal detection analysis; survival analysis (Kaplan-Meier, life tables, Turnbull KM EM estimation, Cox regression, all standard parametric models; type 1, 2, and 3 censoring, etc.); path analysis; Monte Carlo sampling and Markov chainMonte Carlo methods; probability calculator; design of experiments; power analysis; graphics [general, interactiveEDA, histograms, 2D/3D display box and density plots, scatterplots, smothers, quantile and probability plots, maps (U.S., world, etc.)], geographic projection plots, multiplots, icon plots, function plots. Twenty-one statistical procedures allow standard errors to be calculated usingSYSTAT’s bootstrapping capability. This facility is especially useful for models having relatively few observations, or models that are highly unbalanced. SYSTAT’s Monte Carlo capabilities are also useful in such cases.

Changes may be made to graph locations by using point-andclick facilities to alter axis labels, scales, colors, and symbols. Using the Dynamic Explorer component, 3D graphs may be rotated to visually determine the optimal power or log transformation required to normalize data. Graphical capabilities also allow for the comparison of subgroups and for the creation of overlay charts and transform coordinates. Cases may be selected by several means, including “lassoing” across all charts and in the data editor. QuickGraphs automatically graphs results when it is selected. This allows the user quick feedback when engaged in EDA. Version 11 has added several new graph saving formats, including GIF, TIFF, PS, PNG. Charts can be saved as BMP, EPS, EMF, JPG, PICT, WMF, and CGM files.

Although SYSTAT has excellent graphical capabilities, the vendors recommend purchasing SigmaPlot for the finished graphical product. That is, they suggest using SYSTAT for all initial graphing work, but then importing the working graph into SigmaPlot for fine-tuning and presentation. The current version of SigmaPlot is 9.0. I shall be discussing SigmaPlot 9 in a separate review (p. 111, this issue). However, the graphs produced by SYSTAT are comparable to those produced by packages such as SAS, SPSS, StatGraphics, and other major statistical packages. I should also mention that SYSTAT comes with a 434- page Graphics Reference manual that is replete with completely worked-out examples. Examples range from simple histograms and box plots to sophisticated user-specified layouts for multivariate displays and global spherical maps. The graphical capabilities available to users of SYSTAT are simply excellent, and users should not feel that they must purchase SigmaPlot in order to create quality graphs and charts.


2.4 Data Management

When opening the SYSTAT Dialogue Box with the selection DATA.OPEN, the user has the following choices of data formats with which to directly import data into the Data Editor screen. The default import is, of course, SYSTAT, which has import selections for Data (*svd and *sys), Command (*syc and *cmd), and Output (*syo) files. Other import selections are:

• SAS v 9 (*sd2, *sas7bdat)
• SAS Transport (*xpt, *tpt)
• BMDP Text (*.por)
• BMDP Bin (*.sav)
• Minitab v 11 (*mbw)
• Dbase (*.dbf)
• SPSS (*sav)
• ArcView (*.shp)
• Statistica v 5 (*.sta)
• Lotus (*.wk1, *wk2, *wks)
• Stata (*.dta)
• DIF files (*.dif)
• Jmp v 3.2 (*.jmp)
• ASCII (*.txt, *.dat, *.csv)
• StatView (*svd)

• Drag-and-drop variable selection
• Dialog boxes
• Up to 3,200 variables; unlimited cases
• All standard transformations and conversions
• Matrix computations
• Mersenne-Twister random number generator
• BASIC language for data manipulation
• Save output as RTF or HTML, in addition to the above
• Extensive online HTML help
• Command templates, log, files, and interactive entry

I found working with data in SYSTAT to be rather easy. The help screens were very informative and, in particular, the Data Reference manual provided all of the material required to learn how to manipulate data in SYSTAT. The 250-page Data manual gives thorough overviews of each of the primary data management categories. It also provides the user with a number of worked-out examples. It is extremely well written and easy for even novices to use.


2.5 Command Line Language

SYSTAT provides the user with two ways to analyze data; one is by using the tiered Windows Dialog Boxes, the other by old-style command line entry. It appears that the designers of the various Dialog Boxes have provided windows selection choices for nearly all of the allowed SYSTAT options for the particular module being used. But for those who desire handson control over the analysis process, or who wish to possibly extend the standard built-in capabilities of SYSTAT, the package comes with an extensive high level programming language. A 373-page Language Reference manual accompanies the software. It provides all of the language features of SYSTAT and givesworking examples for each major statistical module (factor analysis, ANOVA, and so forth).



Figure 1. Variable scheme for low birthweight data.

Another nice feature of SYSAT is the ability to replay command sequences.Acommand log is created while the user makes selections from the menu and dialog boxes. One may simply save the command script and rerun the identical analysis at a later time. SYSTAT comes with a library of more than 500 examples which are stored as command files. Users can then modify these command files to use with their own data. They may also choose to run them as is to follow along with the examples in the manuals. This feature is a fine learning tool for novice users.