ADDRESS INFORMATION:
Glenn E. Meyer
Department of Psychology
Trinity University
715 Stadium Drive
San Antonio, Texas 78212-7200
Office: 210-736-8323
e-mail: Gmeyer@Trinity.edu
Recent reviews suggest that there is more to statistics that tests of significance, Other considerations are measures of effect size or percent of explained variance, the need for adequate sample sizes to guarantee a respectable degree of statistical power and the need sometimes to compute combine effect sizes or test them for significance (meta-analytic techniques). These needs can be seen as adding insight to a study or an added burden to the student in introductory statistics.While these topics are covered in some introductory statistics and research design texts, there are not easy ways to carry them out. Texts contain lists of formulas. Sample size calculations entail many steps, the use of complicated charts and then more calculations. There can be added nuances for unequal sample sizes or variances, to z transformations based on natural logarithms or other arcane procedures. Having taught introductory statistics for many years and to audiences with some differences in mathematical skills and intrinsic interest in statistics, it is not surprising to me that the average consumer of statistics in psychology may not be too fluent with these techniques. This opinion is echoed in the literature (Cohen, 1977, 1988, 1992; Rosenthal & Rosnow, 1984)There would seem to be a need for an easy way to calculate effects sizes and samples to encourage students and scholarships to use these techniques. While, some might argue that without hand calculation "one doesn't know where the numbers come from", I would disagree, After all, the power charts and calculations are based on already computed tables and sophisticated formulas. A user of such charts does not do the native calculations (Howell, 1992; Keppel, 1991). Similarly, if one uses a program to compute an anova, you must know that modern programs do not use the computational formulas found in the texts because of needs for numerical accuracy. There are computer packages which offer sophisticated options for power analyses . Also some statistics packages have limited options for reported effect sizes from inputed data analyses (Meyer, 1993). While these may be excellent, they are not particularly user friendly and daunting for the novice. With Power & Effect, we have attempted to design an easy utility for students and scholars to use. Our inspiration came from Cohen (1992) who provided a short and easy method to compute sample sizes. He hoped by supplying such information that it would actually be used. Power & Effect 's design was based on the following strategies:
I. A comprehensive set of procedures for the techniques most
commonly used in teaching.
II. An intuitive graphic user interface
III. Extensive navigation, help and tutorial resources
IV. Macintosh and Windows versions
I. Available procedures.
Power & Effect has two major divisions. :
A. Effect size & B: Sample Size .
A. Effect Size: Calculation of effect sizes from inputed statistics.
Also offered in some simple meta-analytic techniques to combine
effect sizes and/or test a set of effect sizes for significant differences.
B. Sample Size : Calculation of sample sizes
Samples sizes can be calculated for a set of statistical techniques if
one can estimate either the values you would expect (means,
standard deviations, correlations, proportion, etc,) or based on an
estimate of effect size.
Techniques offered are
Regression
Anovas
Chi- Square
Proportion from a Population
As an auxiliary function, if one already knows means, standard deviations, frequencies. correlations, or slopes, Power & Effects offers significance tests for several statistics. These are useful techniques not found in some packages but noted in texts. Also, there are useful utilities for examining already calculated statistics from other studies. Those available are:
Pearson's r :
Slope of Regression Lines
Slope1 versus Slope2
II. Graphic User Interface
A significant problem in designing software is user-friendliness. This is especially the case for statistical software where the novice user may not only be unfamiliar with computer usage but with the problem to be solved. Thus, we have adopted two metaphors to make usage easier by taken advantage of the user's knowledge base. This strategy is the successful of the Macintosh Finder and Desktop, Microsoft Windows and other similar products.
A. Effect Size and the Formula Metaphor
We use a hypertext graphic of the formula for the technique itself . Thus, in the example given, the user would click on each formula term and a field would appear. After all terms are entered, the Epsilon Sq. is clicked upon and the result calculated. Extensive error checking, re-entry and correction capabilities are provided, Thus the user is presented with a form similar to the one presented in a standard text.
B. Sample Size and the Calculator Metaphor
We use a calculator metaphor. Users are presented with a panel that duplicates the needed statistical information to calculate sample size. These are hot hypertext and clicking on time opens a field for data entry. Choice of power level and significance level is made by. calculator buttons. Thus, the user is not left pondering what significance level refers to - or what power level to type in. They are prompted. While this may be seem limiting to purists,it aids novices.
An example for Pearson's r would be:
III . Navigation, Help and Tutorials
A. Navigation
Power & Effect contains extensive navigational capacities. It is easier for novices to be trapped in a program and not know where they are, how to move to another place or how to quit the program. On all windows, you will find a row of buttons that enable one to quit, go to a help menu, or the major technique menus, or to a hypertext subject index that allows you to pick a topic and go to it.These capacities are also duplicated in a set of standard Macintosh pull-down or Windows drop-down menus.
B. Help and Tutorials
Power & Effect has available:
1. Animated tutorials of general usage, Power windows use , Effect
windows use, reviews of effect size and power concepts and using
the Index.
2. More complicated procedures have additional instructions available.
3. Examples of common errors and common error messages
4. Some windows (in the Macintosh version) have a "Bubble help"
facility similar to Balloon Help. When activated, cursor movement
brings up information about the window.
IV. Macintosh and Windows versions.
Versions of Power & Effect exist for Macintosh and Windows systems. Each version has successful run on machines capable of running System 7 or Windows 3.0. Testing on myriads of models has not revealed any significant problems or tendencies to crash. Both versions have been under constant testing and modification for a year. It has been used a supplement for undergraduate beginning and advanced psychological statistics classes and a year long sequence of advanced statistics for a Psy.D. program. The undergraduate and undergraduate classes both used SPSS (Unix, Macintosh or PC versions) has their major computational tools along with statistical calculators (Sharp El-5120 recommended). Texts were Howell (1992) for the undergraduates and Keppel (1991) for the graduates. These texts are not lacking in sophistication. Both groups found the interface easy to use and master. They were able, when presented analyses or when using their own computer output, to easily enter data in the program and draw appropriate conclusions.
