|
StatsDirect 1.615 Reviewed by Virginia Fitzpatrick |
|
|
|
|
| Overall scores | |
| Installation | Excellent |
| Learning curve (beginner who can Web surf and word process) |
Fair |
| Technical support | Very good |
| Features | Very good |
| Customizability | Excellent |
| Utility to biologists | Very good |
| Value for money | Excellent |
Overview
StatsDirect is designed to give research workers with limited statistical and programming experience a robust, easy-to-use software package for statistical and graphical analyses of their data. Toward this end, it includes many well-chosen statistical procedures: simple descriptive statistics, most standard parametric and non-parametric univariate tests, nine types of ANOVAs, parametric and non-parametric multivariable regression, seven types of survival analyses, eight types of meta-analyses, and principal components analysis. There is a strong emphasis on the use of confidence intervals and exact methods. All of the functions are supported by detailed examples in the help system built into the software. With the 18 types of graphics procedures provided, one can easily illustrate treatment effects on the data distribution.
StatsDirect was designed for small to medium-sized data sets. It can accommodate 65,000 records and 256 variables per worksheet, and as many worksheets and workbooks open as available system memory will allow.
Available platform |
Microsoft Windows 98, NT, 2000 |
System requirements |
Minimum 16 Mb RAM; 64 Mb recommended |
Test platform |
Dell XPS B800 800 MHz Pentium III running Windows 98 Second Edition, 256 Mb DRAM |
Price |
£99 ($160) |
How Long Did It Take to Learn to Use It Productively?
The executable download file guides installation in a matter of minutes and leaves a distinctive icon on the desktop. Anyone familiar with basic statistical procedures and Excel spreadsheets should be able to perform analyses with StatsDirect immediately. The program includes a simple spreadsheet that can be populated with data either by manual entry or by import from an Excel or text file.
Product Quality
| Ease of installation | Excellent |
| User friendliness | Very good |
| Interface | Menu-driven GUI |
| Intuitiveness of design | Very good |
Customizability
Given that the program is menu driven, there is an impressive amount of customization available for the output. One can easily choose exactly how much detail to include from a multiple-choice list. The output generated is a rich text format (RTF) file that can be edited immediately within StatsDirect or saved and then edited with a standard word processor. Input can be taken from various types of spreadsheet data structures, and there are many types of analyses from which to choose.
Ability to Program in Scripts, Add Extension Modules, etc.
Extensive menus are used for analyses, but log files and programming options are not available.
Ability to Import and Export in Different File Formats
StatsDirect can import Excel files and tab- or comma-delimited text files.
Useful or Unusual Features
StatsDirect uses a menu-driven GUI to act upon data specified in its own spreadsheet. Upon opening any Excel file, an icon appears in the toolbar. When one clicks on the icon, the spreadsheet will have the full functionality of StatsDirect.
StatsDirect provides a ladder plot for the comparison of paired data, which is a valuable tool for analyses of a crossover lab experiment or clinical trial. Although with some programming this can be done with other software, it is not available in Excel and in the other easy graphics packages.
One of the most advanced statistical methods to be included in StatsDirect is the Wei-Lachin generalized log-rank test for multivariate survival data with two main groups. This procedure is helpful in medical research because it tests two groups for median differences in times for several events taken separately, and then for all the same events taken together. The latter test is sometimes more powerful than univariate event tests and could also be useful for avoiding multiple-comparison problems.
Limitations
The program is menu driven and very easy to use. However, you cannot go back and undo part of what you have done - you just have to start over. A multilevel undo feature is planned for future versions of StatsDirect.
Generally the statistical analyses of StatsDirect were consistent with the results I obtained using SAS and Microsoft Excel. However, there were differences between SAS and StatsDirect for the two-sample test for differences in proportions. While both packages used a version of the Fishers Exact test, the results from SAS were more conservative than those from StatsDirect.
Thanks to StatsDirect's excellent online help, it was easy to find an explanation for this difference: StatsDirect calculates an exact two-sided P value for the hypothesis test using a mid-P approach to Fisher's exact test. Since statisticians do not agree universally on when to use mid P, one must be concerned whether this strategy would find favor with the relevant government, publication, or institutional statistician. Therefore, it might be worth documenting the results of a critical analysis the first time it is performed on StatsDirect along with the results of the same analysis performed with a professional standard such as SAS, and being able to explain the reason for divergence in results, if any.
In the StatsDirect version tested for this review, the online documentation for the Mann-Whitney U test stated, "In this analysis we clearly have to accept the null hypothesis that one group does not tend to yield different fitness scores to the other." Usually statisticians claim that there is no evidence to support the hypothesis of difference in treatment. This is a subtle but important philosophical distinction. I have seen neophyte researchers state in their reports that two groups were the same because the statistical test performed did not show a difference. Unfortunately, they were not sure before beginning the study that the sample size was large enough to provide the power to detect a given meaningful difference. However, in the latest update of StatsDirect (1.616), the documentation has been changed to emphasize "unable to reject null hypothesis of no difference" in all relevant sections of the knowledge base.
Because all analyses are menu driven, one does not have a script or log file to document exactly how an analysis was conducted. When many procedures for a particular analysis must be performed repeatedly, menu-driven procedures are more cumbersome to use than command files with programmed statements for the data analyses.
Comparisons with Similar Software
This package is as simple or simpler to use than other statistical packages on the market that I have tried, including SPSS, Systat, S-plus, and Minitab. For the vast array of graphics, parametric and non-parametric procedures it offers, StatsDirect is very reasonably priced.
Technical Support and Documentation
The online help is excellent. The "Statistical Methods Selection" choice in the help menu is designed to guide a nonstatistician toward the appropriate analyses. There is extensive, clearly written documentation and examples for each procedure. Although there is an international telephone number listed on the Web site, customers are urged to use email for technical support. I sent several email messages asking for information and received responses within 30 minutes.
Target Users
This program would be most useful to clinical researchers who have worked with a statistician in planning their experimental design and analyses and who want the tools to investigate their data. Although the program provides extensive guidance on which statistical test to use in a given situation, there might still be problems with the conclusions to be drawn from an experiment, even if the right test is chosen. Online documentation for a software package cannot anticipate difficulties often due to trial design, multiple comparisons, randomization, and premature "peeking" at the data. These difficulties must be overcome from experience working with a statistician or formal training in statistics.
Working through the examples given for each type of statistical test in the online help's "Statistical Methods Selection" would be an excellent exercise for both first-year undergraduate students in statistics and frontline researchers. The needed example data for the statistical tests are all available in one spreadsheet that is included with the program.
As a professional statistician, I found the extensive documentation for each statistical test one of the most useful and unique features of this program. Often, I like to recheck the assumptions and calculations of some of the tests that I have not used for some time, and StatsDirect makes this very easy. StatsDirect also includes a listing of relevant statistical literature for each test, and I would be willing to pay extra if StatsDirect could also put that literature online. That would also be great for students.
Publisher information |
amCode 1 Westbury Barns West End Ashwell, Herts SG7 5PJ United Kingdom Tel/fax: International: +44 1462 743 656 Web site: www.statsdirect.com Online sales: www.statsdirect.com/asp/user.asp |
Pricing structure |
Academic: 1 user £99 ($160) |
Software class |
Data analysis and visualization |
Virginia Fitzpatrick is an independent consultant for several small HMOs and for a customer-research organization.



Digital Frog 2
KaleidaGraph 3.5
NIH Image 1.62
GraphicConverter 3.8
CodeWarrior Release 5
TotalLab 1.00