Systat 10.2: Comprehensive Statistical Analysis Software
By Paul Barrett
Affiliations:
On behalf of : The Australian and New Zealand Journal of Statistics
The Software
SYSTAT is an extensive suite of statistical analysis routines for analysing quantitative data, explicitly written for use on personal computers, and now in its tenth release. It arrived nicely packaged with 6 hardcopy manuals (2000+ pages) and an installation CD and installed with no problems whatsoever. A set of icons was created in a program folder - with links to online help, the pdf format manual files, the standalone ASCII text editor (FEdit), and the program itself. On the review machine (a 2GHz Pentium-4 standalone PC), non-cached load time for SYSTAT was 4.1 seconds in comparison to that for STATISTICA 6 and SPSS 10.07 of 6.5 seconds.
The software package has a very clean user-interface with the screen partitioned into three key areas. The first is an "output organizer", a navigation screen typical of most statistical software nowadays which displays the analyses that are run on a dataset in tree-view. The second is the "output pane" which displays the text and graphics of any results from statistical analyses. The third is a "command" area in which text commands can be typed in order to manipulate the data, or to construct and/or preview a command file; alternatively, it can be designated as a "command logging" region.
The list of Statistical Analysis routines available to a user in this package is significant:
|
- Bootstrapping and Sampling
- Classification and Regression Trees
- Cluster Analysis
- Conjoint Analysis
- Correlations, Similarities, and Distance Measures
- Correspondence Analysis
- Cross-tabulations
- Descriptive Statistics
- Experiment Design
- Discriminant Analysis
- Factor Analysis
- Linear Models 1: Linear Regression
- Linear Models II: Analysis of Variance
- Linear Models III: General Linear Models
- Logistic Regression
- Loglinear Models
- Missing Value Analysis
- Mixed Regression
- Nonlinear Models
|
- Nonparametric Statistics
- Partial Order Scalogram Analysis
- RAMONA Path Analysis
- Perceptual Mapping (MDPREF, Biplot etc.)
- Power Analysis
- Probit Analysis
- Rank Regression
- Ridge Regression
- Set and Canonical Correlation
- Signal Detection Analysis
- Data Smoothing
- Spatial Statistics
- Survival Analysis
- T-tests
- Item Analysis including 1-2 parameter IRT
- Time Series Analysis
- Two Stage Least Squares
- Multidimensional Scaling
|
Getting data into SYSTAT is achieved either by direct keying into a spreadsheet or by the import of stored data in various file formats such as ASCII, SPSS, Excel, SAS, BMDP, ArcView, and ODBC via SQL. Exporting of data and results files is equally impressive with the capability of exporting results directly to HTML and RTF formats, ready for web and word-processor document construction.
Not only does SYSTAT provide accurate and nicely structured quantitative output, it also provides display graphics of an extremely high quality and diversity. In fact, there is a complete manual devoted solely to the Graphics available for displaying data. These graphics exceed those of BMDP, SAS, and SPSS in terms of quality and all-round utility. Virtually every aspect of a graph from text through to axes, points, and lines, can be edited and adjusted. This is an entire graphics package in its own right. The Influence plots, LOWESS, and other data smoothers available to the user within the graphics subsystems are a fine embodiment of exploratory data analysis.
For automation of repetitive analyses and for creating standard analysis procedures, SYSTAT contains its own command language. This consists of meta-commands which run entire analyses from an "optioned" command-line statement, very much like SPSS in form and utility. It also provides a rudimentary procedural customisation language, SYSTAT BASIC. This programming language enables a user to program specific actions on data files, and augment analysis results with custom calculations and graphics. It is similar to the old MS BASIC/GWBASIC language, highly serviceable if a bit limited in these days of object-oriented programming.
Overall, SYSTAT 10 can be recommended as a robust general-purpose statistical analysis and graphical display system that is at least half the price of its nearest competitors. It has also been around for over 14 years now, which has invariably helped make this a very stable and reliable software analysis system. It is an extremely attractive proposition for educational use in literally any area where undergraduate and postgraduate statistical methods are taught. Further, given the advanced specification of many of its routines, it can also serve as a substantive research tool in its own right.
Acknowledgement
My thanks are extended to Systat Software Inc. for providing the review copy of SYSTAT 10.2 |
