|
SOFTWARE REVIEW JACK YURKIEWICZ, Feature Editor, Lubin School of Business, Pace University With this issue of DECISION LINE, it is my great honor to assume the duties of software review editor. I will try to follow the excellent mold carved by Dr. Barry Render, who held this position for the past six years. I have taught statistics, operations research, and operations management for the past fourteen years and my interest in computers goes back to the introduction of the IBM PC more than eleven years ago. This column will focus on evaluating software that is of interest to the majority of our membership. If you wish to become a reviewer, please call me at Pace University, (212) 346-1908. In future issues I will talk a little more about myself and my philosophy of critiquing.
MINITAB, The Student Versionby Jack Yurkiewicz, Pace University This review will look at the student version of MINITAB and conclude our look at commercial or professional statistical software that has been made available in student versions. The software we reviewed, NCSS and Execustat (December/January 1993), and SPSS StudentwarePlus (July 1993), were all DOS products. In the future we will examine the Windows' counterparts. MINITAB's student version is called version 2.1, but it is actually a subset of the professional version 8.2, the last DOS version. The current professional version 9 is a Windows' product, and as yet no student version of that has been announced. MINITAB requires 3 MB of hard disk space and at least one MB of RAM. MINITAB, the company, says that the software is used in over 2,000 schools and by many major corporations. If you are familiar with older versions, you will be pleasantly surprised by this vari- ant. Whereas previous versions were command driven, this MINITAB uses pull-down menus, dialog boxes, and supports a mouse. Indeed, except for the fact that the interface is character-oriented (rath- er than pixel or graphical), the program looks and acts very much like its Macintosh or Windows' counterpart. Users of older version can still use the program in the command mode. Newcomers or con- verts will probably prefer the menu and mouse support. Capabilities The student version is limited to 3,500 data points, so it is quite possible that these data limitations might make MINITAB inappropriate if instructors want students to analyze ``real'' data or to work on sophisticated cases. Since the program is designed for the first and second statistics university course, it does not have any multivariate capabilities. MINITAB can perform descriptive statistics, find inferences from one and two samples, and do regression, analysis of variance, cross tabs, time series, and statistical quality control. It is this last module, quality control, that makes this product stand out from the competition, offering the most comprehensive capabilities in this area. Besides the standard X-bar, R, s, p, c, and u-charts, MINITAB can also make individuals, moving average, exponentially weighted moving average (EWMA), and moving range charts. Also, because this module was designed relatively late in MINITAB's life (version 7), the graphics output here far outshines that found in the other modules. With each generational update, MINITAB introduced additional pixel graphics, but some plots in this current release are still character based. For example, scatter diagrams, histograms, box plots, and control charts are pixel (the latter are especially informative and attractive), while a time series plot is a disap- pointing character graph, with observations not even connected. Histograms are shown as horizontal bars, a minor anomaly. The time series module has only Box-Jenkins techniques. You specify the model and MINITAB will, using an iterative process, estimate the parameters. Having just Box-Jenkins procedures, MINITAB might be an unfortunate choice for that first statistics course that covers forecasting. Students in such a course are more likely to encounter the class of exponential smoothing techniques (Brown, Holt, Winters) than Box-Jenkins methods. Data Entry Keyboard data entry is done on MINITAB's spreadsheet. It is quick and convenient, and is invoked using the ALT-D keystroke combination. Hitting the F10 key, you get a menu that permits changes in the spreadsheet, such as deleting columns or naming variables. It is an ``open'' spreadsheet, meaning you do not have to specify in advance how many variables and observations your data set has. MINITAB will also import ASCII and Lotus numeric files; labels are translated as missing values. This can be cumbersome when importing a Lotus spreadsheet, where the first row tradi- tionally has the variable names. These names are not imported and you must manually enter them after the data is brought into the MINITAB format. Using Lotus' NAMED RANGE procedure overcomes this, but it is still awkward. The program offers a rich set of transformation capabilities. There is a powerful LET command that will allow just about any transformation when combined with the many built-in functions and logical operators. Data manipulation is also good, but there are some minor interface inconsistencies that might force you to keep your manual nearby when you work. For example, you can delete a variable (column) when in the spreadsheet mode (it is a menu choice), but not a row. To do that, you must go to the EDIT CALC Delete Row sequence. Ease of Learning and Use MINITAB's raison d'etre was that it was a statistical program that was easy to learn and use compared to the competition. But a command interface, while speedy, is harder to master than choosing options from the menus available on most new software applications. The current menu system tries to insulate the user, via dialog boxes, from the commands that run the program. Figure 1 shows a typical dialog box. The method does succeed to some extent, but you will never forget that the program looks for and uses those commands. You always see the standard MINITAB MTB prompt, with a blinking cursor expecting you to type something. If you choose items from the menus, MINITAB translates those to the relevant commands (which you see on screen) and the execution takes place. Still, MINITAB can rightfully claim to be an easy program to learn. This is especially true if you have a mouse; you just point and click on your choices. If you do not have a mouse, learning becomes more difficult because you must use a nonintuitive combination of F2, tab, shift-tab, space-bar, arrow, and enter keys to get things done. Since MINITAB was originally designed as an interactive, batch-run statistical program for the mainframe, this personality is still evident even in this version. Each statistical analysis is an independent execution. For example, making a variable control chart for the process mean and dispersion typically involves an X-bar and R chart. For the X-bar chart you must specify the variable to be analyzed and the sub-group size, and any statistical tests you want run. After MINITAB gives you that chart, you must repeat the entire selection process from another dialog box to get the range chart. If you then want to return to the X-bar chart with assignable cause data removed, the entire sequence must be executed again, making things tiresome. In the SPC module, graphs are integrated with text output. In other modules this is not the case. For example, if you want to see a residual plot as part of your regression analysis, you must run the model (using the standard dialog box that specifies variables, output, etc.), save the residuals as part of your data, and then enter the graphics model (with its dialog box) to get the plot. This process of re-execution inhibits the ``playing with'' or examining data that is so important in statistics. Documentation The 624-page book can only be rated as excellent, and is even superior to MINITAB's own fine documentation that comes with the professional version. Most of the book is comprised of fourteen tutorials, each with a unique ``MINITAB at Work'' example from industry. These tutorials not only teach you how to use the soft- ware, but to a small extent teach the statistical concepts as well. Each tutorial concludes with a set of exercises. The second part of the book consists of a reference section. I found it to be among the best accompanying statistical documentation I have ever seen. Summary and Conclusions The student edition of MINITAB is a very good program with different strengths and weaknesses. Beginning students will find MINITAB easy to learn and to use. Graphics are mixed in MINITAB, and it can handle smaller data sets than most of the competition. MINITAB is clearly the program of choice if your course has a large component of statistical quality control, but would be inadequate if your forecasting component features the traditional class of exponential smoothing models in time series. In summary, this series of reviews looked at student versions of SPSS' StudentWare Plus, MINITAB, NCSS and Execustat. These four programs, along with Statistix (from Analytical Software, St. Paul, MN) and Systat (from Course Technology, Cambridge, MA) are the best DOS student statistical programs available. You should choose the product based on the availability of those features you need for your course. Soon to be released are two Windows' student programs, SPSS and Systat, which have the potential to meet or even exceed the high standards set by these excellent DOS products.
MINITAB. Student Version 2.1
|