Next: Determining effect sizes Fixed
Up: Help for ANOVA Power
Previous: Overview
Subsections
The dialog window consists of 4 main sections:
- Top panel
- is used for inputting sample sizes (or numbers of levels),
and for determining whether factor levels are fixed or random.
- Bottom panel
- has a row for each term in the model. Here, you can
vary the effect sizes and observe the power of the test of each effect
at the specified effect size.
- Alpha window
- in the lower right-hand corner
for entering the significance level of the F tests; this
applies to all the tests in the dialog.
- Menu bar
- contains menus for help and reporting results.
If you accessed this document using the Help menu, you have an
example to look at; otherwise, there is an illustration
later on.
The sections described above
contain the following graphical-user-interface (GUI) elements:
- Menus
- These work just like they do in other programs you run.
Many of them open a view window containing information that you can
subsequently copy and paste into, say, an editor.
- Radio buttons
- are used in the top panel
to determine whether a factor's levels are fixed or random. In a replicated
design, the ``Within'' error is treated as a factor and is always random.
- Input windows
- are for entering numbers. Click the mouse in a window
to enable it for input, and type in a value. The number you type in is
not actually entered into computations until you either hit the Enter
key or click the mouse in some other area of the dialog window; once that
is done, the powers of all the tests are updated.
Note: The rightmost windows in the bottom panel display the powers of
the tests. These are output-only windows and cannot be modified by the user.
- Bar-graph displays
- These are scrollbar-like GUI elements that
work in conjunction with their corresponding input windows. They display
values and, if you click somewhere along the center line of the bar,
the value of that parameter (sample size, factor levels, or effect SD)
will change. You may also drag the end of a bar to obtain a continuous update.
(Like the input windows, the bars for the powers of tests are output-only.)
- Numerical scales
- Each family of bars has a numerical scale; it will
automatically change if necessary to make room for the largest value to be displayed.
You may also manually change the scale range, if you wish, by clicking the
mouse on some scale label and dragging it to a new position. For example,
if you click at ``.8'' and drag the mouse to ``.4'', the display will be
redrawn with the ``.8'' value at the new position, approximately
doubling the scale range. Of course, you cannot modify the scale beyond
the limits needed to display the current values.
One thing to be aware of is that some Java implementations are buggy.
In particular, the initial window may display incorrectly, hiding
some of these elements or even being completely blank. You may need to
resize the window to see everything. Early versions of Java (1.0.2
and earlier, Netscape 2.0, ...) have bugs
that make this dialog fall completely apart, with things appearing in the
wrong places. The only cure for the latter problem is to get a more
up-to-date Java-capable web browser.
Of course, there may also be some bugs in the code itself, and you
should report these to the author.
Next: Determining effect sizes Fixed
Up: Help for ANOVA Power
Previous: Overview
Russ Lenth
6/3/1998