Test Platforms

120 MHz Pentium PC with 16 Mb RAM running Windows 95
486/25 MHz PC with 12 Mb RAM running Windows 3.1.

Overview

Arcus QuickStat (Biomedical) for Microsoft Windows provides numerical tools that facilitate experimental design and analysis. Manipulation tools make it easy to transform, rank, or rotate blocks of data. Sample-size calculations improve experimental design, and the comprehensive suite of commonly used statistical analysis tools ensure that the right test is available for experiments.

Unfortunately, Arcus QuickStat comes with a few problems that may leave some users wary. Cosmetic bugs persist in the latest version of the program (build 133), and more serious calculation problems were evident in earlier shipping versions. Also, a number of QuickStat features may be implemented in ways that confuse novice users.

These minor problems aside, Arcus QuickStat offers an impressive collection of data analysis tools. In addition, QuickStat's author, Iain Buchan, provides excellent support for the program from his Web site, and continues to offer free updates that repair user-reported problems. Finally, the demo version, available from http://www.camcode.com or http://www.researchsolutions.com, allows the cautious user to evaluate QuickStat's performance before purchase.

Program Features

Arcus QuickStat provides an organized interface to input data, view results, and access all analysis tools (figure 1). The program uses worksheets for entering and manipulating data, and operates similarly to conventional spreadsheets. Reports are created when data are analyzed or plotted. Users may annotate and export the text portion of these reports, and may print or copy graphics for use in other programs. Navigation through numerous open Worksheet and Report windows is facilitated by tabs displayed directly below the menus (figure 2).

All data transformation and analysis tools are accessible from menus.

Figure 3

The Data menu includes functions to transform, rank, sort, split, combine, and search data. Several other useful functions include the date and time interval calculator, the random number generator, and pairwise comparison of data (figure 3). The Analysis menu includes all of the statistical tests and tools used in data analysis. These features are described in greater detail in the Analysis Tools section below.

QuickStat also allows users to create customized templates for worksheets and reports. For example, one can create a worksheet with headers and formatting specific for the data. This worksheet template file can be saved and used as a starting point for future worksheets. Unfortunately, this worksheet template must become the default template, and all future worksheets will be structured in this way. The same is true for report templates. QuickStat does not allow naming of templates for specific tasks. It replaces the default template with the customized one.

One can "fool" the program into re-creating the original templates by deleting the Arcwin.wt and Arcwin.rt files from the template subdirectory of the Arcus QuickStat installation, but this approach can be disastrous if one deletes the wrong template files. Reports created after conducting statistical tests use specific report template files. For example, the m_unpair.rtt file is used for the parametric, unpaired t-test. Accidentally deleting this template file means Arcus QuickStat is unable to report the results of this analysis.

Handling Data

The program utilizes worksheets for entering, transforming, and analyzing data. Users may enter data in columns, import data from ASCII or Arcus DOS files, or import data by opening Microsoft Excel 4.0 worksheets. Unfortunately, importing ASCII data as tab- or comma-delimited files does not work exactly as expected. Arcus QuickStat consistently imports all of the data except the last column. This glitch occurs when data are exported as text files from QuickStat, Excel, or created manually in a text editor such as Notepad. Regardless of the source of the ASCII file, the import function always omits the last column in the file. Although very inconvenient, one can overcome this problem by manually adding another comma (or tab) after the last column of data in the text file. The program correctly imports these modified files. Fortunately, the Arcus link add-in for Excel 5 and above is available to make any Excel worksheet seamlessly accessible from within QuickStat.

Users have complete control over the appearance of worksheets. One can customize font type, size, color, and attributes from the Format menu, or by clicking on buttons. Columns of data can be italicized, center justified, or converted to percent by simply clicking on a button. Unfortunately, this flexibility leads to a few problems when working with large fonts: users can increase the size of fonts so much that the numbers become too big to be viewed in the worksheet. If the entries are too wide for the columns, they are displayed as # rather than the actual number entered. Although users can choose Automatic Column Width from the Format menu to correct this problem, there is no way to globally increase the height of each row to display the entire value entered.

Data transformation tools include log, logit, probit, angular, and cumulative, among others. Unfortunately, QuickStat's presentation of the results may confuse users.

Figure 4

First, the program displaces data one row down when the results are saved in the same worksheet. Arcus QuickStat enters the new column name in row 1 of the worksheet (figure 4), rather than in the header, where all of the original column names are located.

Second, when QuickStat performs a log transform, the program really does a natural log transform (usually represented as "ln") rather than a base-10 log transform (conventionally represented as "log"). To do a base-10 or common log transform in QuickStat, one cannot use the built-in transformation tools, but must use the Data menu's Apply Function, and apply the function "clog." This problem is confusing until recognized by the user.

Analysis Tools

Figure 5

QuickStat offers an impressive number of statistical tests and analysis tools. In addition to simple descriptive statistics (figure 5), QuickStat includes exact tests, chi-square tests, tests on proportions, sample size estimation, and probability distribution. Exact tests include Sign, Fisher's exact, Fisher's expanded exact, and matched pairs exact tests. Users can choose from 2x2, McNemar, Mantel-Haenszel, or Woolf chi-square tests, and single, paired, or unpaired tests for proportions. One can conduct estimation of sample size for paired and unpaired t-tests, independent and matched case-control, or independent and pair prospective studies, as well as population surveys.

The program includes a number of parametric and nonparametric statistical methods. Parametric tests include the paired t-test, unpaired t-test, single sample t-test, F-test, and Shapiro-Wilk W test. Nonparametric methods include the Mann-Whitney, Spearman's rank correlation, Kendall's rank correlation, Kruskal-Wallis test, Friedman test, Cuzick's test for trend, two-sample Smirnov test, chi-square (goodness of fit) test, and the Wilcoxon signed rank test.

Analysis of variance (ANOVA), using both parametric and nonparametric methods, is also available, along with the appropriate post-tests for multiple comparisons. ANOVA methods include one-way, two-way, two-way with replicates, two-way nested, crossover, and Latin square ANOVA. Kruskal-Wallis and Friedman nonparametric tests are also provided, as well as post-tests such as Bonferroni's, Tukey-Kramer, Dunnet's, Newman-Keuls, and Scheffe's.

QuickStat also includes tools for regression and correlation analysis. Figure 6 shows the results from a simple linear regression. Users may also conduct multiple linear regression, grouped linear regression, and polynomial regression. One can choose to fit polynomials with 2 through 9 coefficients, and can plot fitted lines with 95% confidence intervals (figure 7). Other regression and correlation tools include logistic regression, probit analysis, and principal components analysis. Probit analysis enables one to conduct dose-response curve fitting of quantal data with binomial error distributions.

Arcus QuickStat provides a number of methods for analyzing survival. The tools available are Kaplan-Meier survival estimates, simple life table, log-rank, and Wilcoxon tests, as well as the Wei-Lachin test. Figure 8 shows the survival plot created after the completed Kaplan-Meier analysis. In addition, Arcus QuickStat automatically creates a hazard plot, a log hazard plot, a lognormal survival plot, and a hazard rate plot.

Graphics

The graphics capabilities of QuickStat are functional, but somewhat limited. The plot types supported include box and whisker (figure 9), cochrane, histogram, ladder, normal, scatter, survival, spread, and ROC plots. Notably lacking is the ability to generate simple bar or column graphs with error bars. Users can copy plots to the Windows clipboard for use in other programs such as Microsoft Word or PowerPoint; when printed to high resolution printers, these are adequate for publication and presentation.

Unfortunately, users have little control over plot attributes. One must select options prior to creating a graph, and cannot modify plots once they are created. The attributes that can be changed include the fonts for titles and labels, the style and color of symbols, and the width and style of lines used to connect symbols.

Arcus QuickStat automatically creates ASCII text versions of histograms, scatter, and box and whisker plots. This handy feature makes it easy to include graphs in email messages, and offers an option for users who want to export ASCII plots with the rest of the text contained in a report file.

Any Problems?

Arcus QuickStat (Biomedical) is plagued by a handful of program bugs or poorly implemented features. Most of these problems, being simply cosmetic, are therefore harmless. For example, figure 10 shows an example of a dialog box that can be resized, leading to odd placement of buttons. Another minor problem arises when one conducts sample size calculations. Proportions can be compared using independent prospective or independent case control. When selecting the former, the program gives calculations correctly in the final statement presented to users.

Figure 11

When selecting independent case control, the values are not placed into the text, and fields for the variables are displayed instead (see figure 11).

The initial shipping version of Arcus QuickStat received for review contained a few problems, but these were fixed in the latest update downloaded and installed from the author's Web site. For example, Ctrl-insert and Shift-insert were used for copy and paste, rather than Ctrl-C and Ctrl-V. But this reviewer observed a more serious problem when conducting ANOVA on four columns of data: the results were simply incorrect, reporting no significant difference (P=1) when the differences were actually highly significant (P<0.0001). Although this is not a problem in the current version, it does emphasize the importance of obtaining the free updates as the author makes them available.

Documentation and Support

The printed documentation is an excellent guide to using Arcus QuickStat. Features are clearly explained, and the examples provided for each statistical test serve as a tutorial for the program as well as the method being applied. Most helpful are the "inference" statements that follow each section; these statements are essentially an interpretation of the statistical test applied, and facilitate greater understanding of the methods.

Of course, this evaluation uncovered a few minor errors. Arcus is spelled "Arucs" on page 77 of the user's manual, and a hypertext link to "Excel links" in the online help file leads to a box with no additional information. Fortunately, Buchan is responsive to concerns expressed by Arcus QuickStat users. He maintains a discussion list on his Web site, where he continues to make updates to the program freely available.

The Bottom Line

Arcus QuickStat (Biomedical) provides researchers with a comprehensive suite of data analysis tools that facilitate experimental design and analysis. Anyone familiar with spreadsheets will find that the QuickStat worksheets offer flexibility and power when entering and manipulating data. The extensive list of analysis features ensures the correct test is available for statistical analysis of data. Sample size calculation, an essential feature for any biomedical researcher involved in experimental design, is well implemented and easy to use. The graphics capabilities of QuickStat are functional, although a bit inflexible. The ability to create ASCII versions of many plot types will be a pleasant surprise for researchers who must send text plots via email. Unfortunately, Arcus QuickStat has a few implementation problems that may leave potential users wary.

Figure 12

This review omits many of QuickStat's other features. For example, users can set up external applications on the Tools menu of QuickStat. This enables one to launch Web browsers, email programs, or applications such as Notepad from within the program (figure 12). The Arcus Calculator, an application included with QuickStat, offers an alternative to the Windows 95 calculator. Finally, the review has ignored a number of statistical tests. To learn more about Arcus QuickStat or to download the demo version of the program, visit Research Solutions Web site, or the author's Web site.

System Requirements

Arcus QuickStat (Biomedical) is available for Microsoft Windows 3.1 or higher, running on a 386, 486 or Pentium computer, with a minimum of 8 Mb RAM, and 3 Mb hard disk space. Free updates to the program are available to registered users.

Purchasing Information

Arcus QuickStat is available from Research Solutions, which can be reached by email, by phone at (01223) 425 558 (in the UK) or +44 (0) (1223) 425 558 (outside the UK), by fax at +44 (0) (1223) 425 349, or by mail at 124 Cambridge Science Park, Milton Road, Cambridge CB4 4ZS, United Kingdom.