The illustrations in this article are intended to show some of what is possible in the way of providing a reasonable GUI for working with power and sample size. The key objectives are to make the interface as straightforward as possible, and to keep it as uncluttered and nontechnical as possible.
The Excel add-in has the advantage of relying on widely available commercial software and being integrable in many ways into other work that is being done in Excel (and in other applications, via dynamic data exchange). Workbooks designed for one scenario are easily modifiable or extended to others.
A typical spreadsheet application relies on formula cells and data cells. To get a cleaner, less cluttered interface like in Figure 3 we need cells to serve both roles, requiring programming as extensive in Excel as in Java. Excel's Visual Basic language is a great improvement over its previous macro language, but it is still not as powerful or extensible as would be desirable; for example, the VB in Excel version 7.0 is object-oriented, but it does not provide a means of creating new objects.
Java applets are harder to develop than spreadsheet applications such as Figure 2, but the advantage is a powerful, portable language that makes it possible to build GUIs for complicated situations. By taking advantage of parameter tags, one can develop general-purpose applets. It is possible for a statistical consultant to write a script like the one in Figure 5 and then to provide it to a researcher over the Internet.
There is much more that can be done. Obviously, interfaces need to be created for many other statistical tests than are covered here. Even within the context of analysis of variance, for example, our Java applet could be improved by providing for omitting some terms from the model, and unbalanced and fractional designs. We could also benefit from more flexible effect-size specification or illustration---perhaps through a ``wizard''-style interface.