A Review of Eight Statistics Software Packages for General Use

Article excerpt

1. INTRODUCTION

There is a great need, in scientific research and in the business community, for statistical software packages that enable nonprofessional statisticians to accurately and efficiently evaluate data. The data being analyzed may be archival or the results of a carefully planned study directed specifically toward current questions. In any case, the intelligent collection and evaluation of data, using generally accepted techniques, is critical for effective process understanding and decision making.

With the near-exponential growth of PC computing power, many statistical techniques are available at the desktop, provided by software packages that cover a wide range of analyses and statistical graphics. Eight software packages are evaluated in this review. These packages are provided for use by those who are not professional statisticians, but who have need for statistical treatment and evaluation of data. With this orientation in mind, the packages in this review have been evaluated on such features as quick start-up, interface design, usefulness and clarity of help information, and clarity and accuracy of results, as well as breadth of capability.

The software packages herein reviewed are GraphPad Prism, InStat, ISP, NCSS, SigmaStat, Statistix, Statmost, and Winks. These products have all had wide acceptance in the market and appear to have a base of loyal customers. Each have strengths and weaknesses, and each differ in their user interface, breadth of analysis, level of statistical sophistication, area of focus, and price. Information is provided here to help the reader better understand some of these tradeoffs, especially when making recommendations for colleagues.

The review process I maintained for all packages consisted of completion of any tutorials provided with the software, followed by analyses of a set of some 20 test problems. Results were compared to those obtained using SAS software (SAS Institute) for accuracy. Any accuracy problems found are included in each program review. I had no previous experience with any of these products prior to this review. In the following section, summary tables of statistical capabilities and ease of use are presented, followed by overviews of each package that focus on system requirements, data import capabilities, user interface, and statistical capabilities (breadth and accuracy). Recommendations are presented in Section 3. Recognizing that these packages have ongoing development and frequent upgrades, some changes that address deficiencies discussed here are presented in Section 4. Vendor information and package prices are given in the Appendix.

[TABULAR DATA FOR TABLE 1 OMITTED]

2. PROGRAM FEATURES AND DISCUSSION

For this review, 23 sample data sets were stored in a Microsoft Excel workbook. In general, it was easiest to get these data into the statistical packages by using the Windows clipboard. This approach was very straightforward and convenient, since all the packages display data in a spreadsheet format, providing a very intuitive transition. In some cases data were reformatted before copying, but this was still generally a simple matter. I would recommend that users of any of these statistical packages have a full-featured spreadsheet program available to facilitate data manipulation for preparing data sets. Frequently, data for analysis come in unusual formats, requiring transposition or, perhaps, removal of interspersed summary statistics or other notes. Most of these editing activities are more easily handled in a spreadsheet program rather than inside one of the statistical packages.

In the following discussion, I provide summary tables [TABULAR DATA FOR TABLE 2 OMITTED] that present most of the statistical capabilities and usability features of each of the programs. These tables are not completely enumerative, but they do include all the major areas of analysis capabilities. The tables are followed by discussions of each of the packages. …