Summary

What G*Power can and cannot do for you

G*Power is a completely interactive, menu-driven program for IBM-compatible and Apple Macintosh personal computers. It performs high-precision statistical power analyses for the most common statistical tests in behavioral research, that is, t tests, F tests, and chi-square tests. G*Power computes power values for given total sample sizes, effect sizes and alpha levels (post-hoc power analyses) sample sizes for given effect sizes, alpha levels, and power values (a priori power analyses) alpha and beta values for given sample sizes, effect sizes and b/a ratios (compromise power analyses). The program may be used to display graphically the relation between any two of the relevant variables and it offers the opportunity to compute the effect size measures from basic parameters defining the alternative hypothesis. This article delineates reasons for the development of G*Power, and it describes the program's capabilities and handling.

Why G*Power?

G*Power was designed to serve as an efficient, broadly applicable and easy-to-use research tool. Therefore, options which are useful primarily in an educational context (e.g., Monte Carlo simulations or illustrations of the relation between mean differences, error variances, and effect sizes) were omitted. Good programs for these purposes have already become available (e.g., Borenstein & Cohen, 1988; Borenstein, Cohen, Rothstein, Pollack, & Kane, 1990, 1992; Rothstein, Borenstein, Cohen, & Pollack, 1990). In developing G*Power, we gave priority to providing for a variety of power analyses (a priori, post-hoc, and compromise power analyses) for most of the common statistical tests in behavioral research. It appears that t tests (see t-Test on Means, t-Test on Correlations, and Other t-tests), F tests (F-Test (ANOVA), F-Test (MCR), and Other F-Tests), and Chi-Square Tests characterize this class sufficiently. Moreover, we aimed at high-precision power calculations which are offered by only a few of the available power programs (see Goldstein, 1989). A high level of precision is especially important for power analyses based on small alpha and beta values (as they occur, for instance, when alpha or beta are adjusted in order to control for the cumulation of error probabilities; see Westermann & Hager, 1986).

How to Handle the Program

The present versions of G*Power assume that users are familiar with the basic concepts of statistical power analyses. Moreover, it is useful if users know about Cohen's effect size measures and the definition of "small," "medium," and "large" effect sizes (see effect size conventions). The relevant background information may be found in Cohen (1988). Apart from that G*Power also allows to calculate the effect size measures from basic parameters such as means, variances, and probabilities. [...]

Hardware and Software Requirements

The Macintosh version of G*Power should run on any 68K Macintosh using system 6.0.7 or higher. [...]. Two different implementations are available, one that requires and takes advantage of an arithmetic co-processor (i.e., G*Power/FPU), and one that does not (i.e., G*Power). There is also a more recent version which runs natively on Power Macintosh computers (very fast!). The MS-DOS G*Power version requires an IBM-compatible PC with MS-DOS 3.31 or higher and a graphic card. G*Power may also be used in the DOS windows of WINDOWS 3.1 or OS/2 2.0. We recommend to install the program on a 386 (or better) PC with an arithmetic co-processor. To take full advantage of all G*Power options, a VGA graphic card and a color monitor are necessary. A mouse is not necessary but very helpful. When using the MS-DOS version without a mouse one selects options by pressing the key corresponding to the appropriate highlighted letter (selecting items within the active region of the window) or by pressing "Alt" plus the key matching the appropriate highlighted letter (activating another region of the window and selecting an item from the new active region). If the same letter is highlighted twice, it is always in different regions of the window. Within parts of the regions, items can also be selected with the arrow keys.

Availability of the Program

G*Power may be obtained free of charge. The most convenient way to get a copy of G*Power is to download the program via our G*Power web page. Alternatively, Macintosh users may download "gpower2.sit" from the "MacPsych archive for psychology concerning the Macintosh computer" (see Huff & Sobiloff, 1993, for details) Transfer of the programs via regular mail is also possible. Interested readers should write to E. Erdfelder or F. Faul if they want the MS-DOS version, and to A. Buchner if they want the Macintosh version. A completely new, unformatted floppy disc must be enclosed. In publications involving the use of G*Power, we kindly ask you to cite the program version used.

Warning

Although considerable effort has been directed toward making G*Power error-free, there is no warranty whatsoever. Users are kindly asked to communicate problems encountered with the program to the authors.

Please report suggestions for improvements to Axel Buchner, Franz Faul, or Edgar Erdfelder.