Emerald Group Publishing Limited
Copyright © 2008, Emerald Group Publishing Limited
Article Type: Computer currency From: Journal of Consumer Marketing, Volume 25, Issue 4
Edited by Dennis A. PittaUniversity of Baltimore
In this issue we are privileged to have a review of a mature update of one of the premier pieces of statistical programs, Systat 12.0.
Several years ago, statistical applications were limited for the use of the specialists; now with greater availability of graphical software within easy reach of the computer user, it is becoming easier and easier to use statistics. Systat 12 brings us a step closer to using basic as well as advanced statistical tests. Considering various statistical tests that a survey researcher is likely to use, Systat is amongst the major software programs such as SPSS or SAS.
The program comes in a single CD. The installation is quick and without any problems. The CD offers an option to read the installation instructions. For those needing installation instructions or an overview of the process prior to installation, it is a good idea to click on this icon. I started directly by clicking the “install Systat 12” icon and chose various options as they became available.
Once you start the program, it opens up three major windowpanes. The top ribbon shows various menu items like any other windows program. To the left is a vertical pane called workspace. This shows the major headings of output sections. The major portion of the screen shows data window, graph window or the output window, depending on the usage and or selection by the user. The third pane or the lower section is for running interactive commands or viewing the log of executed commands. For those familiar with any other statistical program, the interface is very intuitive. For those not familiar with such programs, I found the tutorial under help menu to be extremely useful.
The interactive options in commandspace work very well. As the user becomes more familiar with the program and the interactive commands, one can use the commandspace just as or more effectively than the graphical interface. For example, just typing CORR (abbreviation for correlations), followed by PEARSON prints out the correlations of all variables. Typing help in commandspace brings out the help window.
The default view of Systat shows three main windows: the data window (appearance of a spread sheet), the work window (shows major headings of output), and syntax or command window. The data window has further views - data or variable view, whereas the command Window has three views - interactive, log and untitled.
There are several useful customization features. One of them is the “theme.” The theme is not just for selecting a color scheme; it changes the items in the main ribbon or the menu items. For example, the default menu items are: file, edit, view, data, utilities, graph, analyze, advanced, quick access, addons, window, and help. Each of these menu items has several action items. This default theme can be changed under the menu of utilities, and then themes. The theme changes the menu items. For example, in introductory statistics theme the main ribbon menu items are: file, edit, view, output, data, distributions, and relationships.
The data editor window has the familiar look of a spreadsheet. It is similar to Excel with the added feature of “variable view” window which is similar to SPSS. Here one can assign, and edit the variable labels, variable type (string, numeric) and value labels. The full variable label is visible by placing the cursor over the variable name. It is possible to open more than one data file. One user-friendly feature is that the background color of data screen is different for each data file. As a footnote comment, if you click on Systat program through the start, program button in windows while one program is running, it opens another Systat program instead of indicating something like, “the program is already running,” which can be a beneficial feature depending on user’s perspective.
The file, open, data command can open (import) files in different formats such as Excel, SAS, SPSS, Minitab, Statview, Stata, Statistica, and many more. The merge function is very easy to use. Data files can be merged to aggregate data by adding cases or variables. In order for merge function (adding cases) to succeed the two files must be similar in terms of number of variables, name, type (string or numeric), and order of variables.
One of the basic concepts in statistics is the probability density function and the area under the curve. Although basic, it is not easy to communicate or explain to students. The graphical interface of statistical programs makes it easy to calculate the coefficients, the corresponding test ratios such as t-ratios or chi-square and the resulting significance. Students learn simple rules such as p<0.05 is significant, but often they fail to grasp the underpinnings.
The probability calculator helps to overcome this limitation. Under utilities menu, click on the probability calculator, select the appropriate distribution, and specify the values and it calculates the p-value. More than that it also shows the graph of density function (and cumulative density function), so students can see the meaning behind the p-value.
Random sampling can be used to generate a number of samples of given sample sizes of various distributions. I tested the option by generating 20 random samples of size (n) 20 with a mean of zero and standard deviation of one. I then used the graph function to generate one histogram showing overlays of all 20 samples. This can be a useful device to show students the idea of random sampling in survey research and to convey the point of statistical significance and confidence intervals.
Systat has several widely used graphical functions such as histogram, scatter plot, density functions, line chart, bar chart. The convenient feature is the ability to change or modify the graph through the interactivity feature. For example, one can start with a scatter plot and then fit various lines of best fit including showing residuals and confidence intervals. In managerial applications of survey research the analyst may want to show the mean values of a variable plotted against time (X axis) or another independent variable. This can be done easily and the graph can be “cut and pasted” into another document. Most of the programs I could execute without reading the manual. However, for some graphical applications and data manipulations I found that reading the manual was very helpful.
In addition to the commonly used statistical tests such as descriptive statistics, regression, correlation, discriminant, cluster, factor analyses, etc., the analyze command also includes time series analyses. Furthermore the time series includes several types of time series analyses, e.g. autocorrelation, partial auto correlation, cross correlation, trend analyses, Fourier series, ARIMA, etc. The analyze command also contains several non-parametric tests such as Friedman, Kruskal-Wallis, Wilcox, Sign and many more. It is one of the more complete statistical set of options.
The advanced statistics menu option includes more specialized programs. Amongst these are: perceptual mapping software such as perceptual mapping and multi dimensional scaling, and special statistics. Other programs are conjoint, structural equation modeling (Ramona), test item analyses, response surface analyses, survival analyses, trees (multiple classification analyses), and missing value analyses. Several of the items mentioned under the advanced statistics can be found as self-standing software from competing software suppliers. For example, various Conjoint programs are available from Sawtooth Software (www.sawtoothsoftware.com) amongst others, missing value analyses (stand alone from SPSS.com), path analyses (Lisrel from Scientific Software), and so on. Thus, all in all, Systat provides a comprehensive set of tools.
The software has all (in my usage) statistical programs and also some additional unique features. One of these was the probability calculator (mentioned earlier). Some others less commonly included programs are under the regression option. A couple of these are partial least squares (PLS) and ridge regression. PLS is a specialized option that is often a stand-alone program instead of one you can find in software packages. Ridge regression is a very useful procedure when using linear regression under moderate to high multicollinearity. I have used SPSS ridge regression often. In SPSS, it is an additional script not part of the graphical menu interface that needs to be used to estimate ridge trace and ridge coefficients. In contrast, the ridge option in Systat is seamless and works like any other graphical interface option.
My experience in using Systat
I use the analysis of variance often, so I tried the application. It just took a few clicks to specify the dependent and independent variables. There were other options but I stayed with the default options. The output shows the variables in the file, variable codes (independent variables), the multiple R-value, estimates of effects (B coefficients), The ANOVA table, cell means, and overall mean values for various levels of variables. The output also includes applicable graphs showing overall means and cell means, and Durban Watson statistic to check for Auto correlation. One feature I liked about the graph was that they show the error bars (plus or minus 2 standard errors) so one can visualize the significant differences in mean values. At this stage of estimation, the program does not automatically generate post hoc tests. These have to be performed as a separate test after the initial estimation. It just takes a few extra clicks; it is not necessary to re-estimate the model.
In marketing research one often needs to compute the reliability of the scale. The advanced (statistics) menu shows the test item analysis command that has two options for computing the scale reliability: odd-even or split half. The Chronbach’s alpha was not obvious from the menu. I had to look up under help and search. The command was under analyze, correlations, Chronbach’s alpha. It worked fine, and gave me the alpha coefficient. Unlike SPSS, it does not have the option of providing additional statistics such as correlation matrix, and alpha if item deleted. But these coefficients can be obtained by clicking the mouse a few times.
I found the power of the test to be a useful and easy feature to use. In hypothesis testing one discusses the type I and type II errors. Most studies mention type I error or alpha and skip discussing beta (type II error), or power of the test (1-Beta). I found it easy to calculate the sample size needed for a given power or alternatively to calculate the power curve for various sample sizes.
I tried both hierarchical cluster analysis and K-means (there is also a K-median) cluster analysis. Both gave me the output that was quick, simple, and meaningful. I also tried various graphics programs. These were also easy to use and gave the desired results. In survey research the recode command is a must to reverse code the scale values. Systat offers a transform command under data menu and can handle most types of recoding or transformations. I tried some recoding of variables and also created a new variable with “let” command. It worked smoothly.
The help section is useful and worth running on line. The help menu has separate categories for help on Monte Carlo, quality analysis. The quality analysis is an add-on program that provides quality control and analysis tools like gauge R&R studies, sigma measurements, Taguchi’s loss function, and Taguchi’s online control with the beta correction technique. These features provide additional analyses beyond traditional quality control charts and graphics tool base module of Systat.
One feature I found particularly helpful which would be appreciated by new users especially is the tutorial. It shows use of the program in a video format with actual examples. The video shows the mouse arrow clicking key menu items and the corresponding result of these executions.
When you install the program, it offers the option to install all manuals in PDF form. These manuals can be accessed through the Windows program directory. Although the help items, the tutorials and the manuals are all very good, those familiar with other statistical programs should not have any difficulty in using Systat. I am very familiar with the SPSS and other Windows program interfaces and found that transferring from one program to the other was without any incidence
Technical support or contact Systat
Technical support is available through help under main menu, Systat web site (www.systat.com). There is discussion forum also. In addition, Systat support representatives can be reached via e-mail or telephone. I called the technical support twice. Both the times, the phone was answered quickly. In one case, the tech support gave an answer instantly and in the other case, the support person did some research and got back to me via email on the same day. I was very pleased with the help.
Available from: Systat Software, Inc., 1735, Technology Drive, Ste 430, San Jose, CA 95110, USA. Web site: www.systat.com; Sales: 800-797-7401; Fax: 800-797-7406; E-mail: firstname.lastname@example.org Support: email@example.com
Considering the number of statistical tests, both the $1,299 commercial price, and the $999 price for a single user government or academic license are reasonable.
Overall, I like the software. It is easy to use and has all the functionality I need. It also has some unique features not readily found in other packages and is very reasonably priced. I am considering using it for my classes. If I were to buy statistical software, Systat would be on top of my list. Considering the number of different tests it can perform, and the price, it is a great product.
This review is not meant to be a systematic comparison of SPSS and Systat, rather a few anecdotal comments based on my usage of these programs. Of course, each software has its own unique strengths and shortcomings.