The arrival of the first mainframe software and then their PC-based successors has revolutionized multivariate statistical analysis.  No longer are analyses limited in size or complexity and the full advantages of these techniques can be utilized.  However, even a decade ago, the availability of this software was quite limited and even the available software often only performed basic analyses.  But now the emergence of the PC and the recognized viability of commercial multivariate software packages have made these techniques available to everyone.


The software listed below is just a sample of the total packages available, but they have been selected as to what we feel are the best examples of multivariate statistical packages.  There are three categories:


  1. Complete Multivariate Packages -- contain all of the multivariate techniques and more (typically enterprise and/or workstation capable)
  2. Basic Multivariate Packages -- contain the most common multivariate techniques, but usually not all of them
  3. Specialized Statistical Methods -- address a particular statistical analysis (e.g., SEM or time series) in more detail than the more comprehensive packages.

Please feel free to send us your comments or let us know of a statistical package for listing in any of these categories.


Complete Multivariate Packages

    These statistical packages provide a complete suite of not only the basic multivariate techniques (factor analysis, regression, discriminant analysis, ANOVA/MANOVA, conjoint analysis, canonical correlation, cluster analysis and multidimensional scaling), but many of their variants and extensions.  Several of these programs have backgrounds as mainframe applications oriented toward academic research, although several have been developed specifically for the PC environment.

Perhaps the most "educator-friendly" software package, Minitab has been included in numerous textbooks as the accompanying software. Contains all of the basic multivariate techniques along with some specialized topics (e.g., censored data models, time series analysis). Fully functional  demo version available for 30 day trial.
One of two most widely known statistical packages (along with SPSS), SAS has a long history in academic and commercial research, first in the mainframe environment and then migrating to the PC platform. It acts as a programming language and users have developed a large number of specialized routines. The website provides information on the SAS Institute, it's products and services and an excellent set of statistical resources (examples, FAQs, etc.).

The second of today's major software packages (along with SAS) that had its origins in mainframe applications, but has moved aggressively into the PC environment.  SPSS has a full range of techniques, even including a complete conjoint analysis package and structural equation modeling program.  SPSS is also rapidly making associations with numerous additional statistical techniques, ranging from text-analysis to structural equation modeling.  Provides an informative online newsletter SPSS Direct, which addresses not only product information, but basic statistical issues as well.


One of the statistical packages developed solely in the PC environment, the Multivariate Methods module contains all of the basic techniques.  The program is heavily oriented toward graphical analysis and display.


Perhaps the most comprehensive of the statistical packages originally developed exclusively for the PC, Statistica addresses all of the basic multivariate techniques and then some, even structural equation modeling.  Its orientation is also towards graphical display and output.  Demo available.

Contains all of the basic multivariate techniques (except MDS).  Operates in a Windows environment and relies heavily on graphical display and portrayal of results.
Basic Multivariate Packages

These programs perform the basic multivariate techniques (regression, ANOVA/MANOVA and usually factor analysis), but vary in their inclusion of the other techniques (e.g., discriminant analysis, cluster analysis, MDS or conjoint).  Many of these packages may be oriented towards the basic statistics courses.


Developed by Paul Velleman, this program's uniqueness is its exploratory data analysis through both graphical and analytical means.  Originally developed for the Macintosh, it makes excellent use of graphical displays to assess relationships in your data. Modules are also available as Excel add-ins and for data visualization.


NCSS (Number Cruncher Statistical System), while limited in the multivariate techniques included, has been designed to function especially well in classroom environments. There is a  freeware version suitable for teaching or relatively simple data analysis.


    One of a new generation of statistical software, Rweb is a forms-based approach to multivariate analysis that occurs totally over the Internet.  Enter your data via a form (or upload a data file), and then perform Principal Components Analysis, Correspondence Analysis, or Multiple Correspondence Analysis.  Results are computed online and returned to the browser.
    Combining SIMSTAT (regression and ANOVA, with resampling capabilities) and MVSP (MultiVariate Statistical Program) provides the basic multivariate techniques.  Additional modules for item analysis (STATITEM) and factor analysis (EASY FACTOR) are available.
    More oriented toward life sciences, but gaining popularity in many disciplines, Stata has an emphasis on regression and its variants, ANOVA and its extensions and factor analysis.
    StatMost contains most of the multivariate techniques, but its uniqueness comes from its extensive use of graphical results to assist in both exploratory analysis as well as an excellent means of portraying the results.  DataMost also has two other complementary statistical packages,  Numerica (numerical analysis), and DataTrix (a spreadsheet-type package).

    Developed by Forrest Young, ViSta (Visual Statistics) is a free statistical package that includes regression, principal components, MDS and correspondence analysis. It was developed around a four step process in data analysis: reveal structure in your data; provide guidance; show results; and structure the data analysis procedures. Available for Mac, Windows, and Unix.



Specialized Statistical Methods


This final set of statistical programs represent the specialized software directed at a limited range of analytical issues, but with more analytical detail and complexity than can be afforded in the more general programs.  These programs play an important role in augmenting the more comprehensive packages and providing access to emerging techniques before they are incorporated into the more general programs.




    This structural equation modeling package by James Arbuckle is becoming quite popular due to its ease of use and integration with SPSS.  A free student version is also available.


    EViews is an econometrics package with an emphasis on econometrics estimation (single and multiple equation systems, all types of dependent variables), forecasting and simulation.


    The GAUSS Mathematical and Statistical System  is a high level matrix programming language specializing in commands, functions, and procedures for data analysis and statistical applications.



    LIMDEP is an econometrics program for linear and nonlinear models, limited and qualitative dependent variable models, models for cross section, time series and panel data.


    LISREL is perhaps the most widely used structural equation modeling (SEM) program.  As one of the "original" SEM packages, it has defined the nomenclature for specifying structural equation models.  It has evolved into a general-purpose latent variable modeling program, with extensions to address such model forms as multi-level, growth curves and time series.


    RATS is an econometric time series software for analyzing time series and cross sectional data and developing and estimating econometric forecasting models.


    SORITEC is an econometrics and statistical package for estimation of all forms of dependence relationships, including discriminant analysis and variants of regression and other linear models.  It does not address experimental design analyses.


    These are techniques based on exact tests (or permutation tests), typically used with small-sample categorical data.  StatXact is for general permutation tests with small-sample categorical data and  LogXact for exact small-sample logistic regression.



Drop us an e-mail if you have a comment, suggestion
or online resource you would like to share?


Multivariate Data Analysis
Hair, Black, Babin and Anderson