Information Technology Reference
In-Depth Information
Figure 10.23
The
Random Effects
screen after clicking
Add
.
levels. Subjects in a research study are examples of such variables. We sam-
plethemasbestaswecan,randomlyifpossible,butwewishtogeneralize
beyond our participants to the population they represent. Random effects
are contrasted with
fixed effects
, where we have specifically selected certain
levels or conditions of an independent variable so that we can describe
their effect with much less concern about generalizing to other conditions
that are not included in the study. This topic is discussed in Chapter 17.
More extensive discussions of random effects may be found in Maxwell
and Delaney (2000) and Keppel and Wickens (2004).
Although SPSS did not broach this issue in its
GLM
procedure,
SAS
Enterprise Guide
requires that we treat subjects as a random effect when
we are using within-subjects designs. When independent variables are
treated as random effects in an ANOVA, different error terms are used
to compute the
F
ratios in the study. As a consequence of this, at least
for the within-subjects designs we describe in this chapter and in Chap-
ters 11 and 12, the
F
ratios produced by
SAS Enterprise Guide
will be
different from those produced by SPSS.
To specify our subject identifier as a random effect, select
Random
effects
under the
Randomeffectsandoptions
frame; a dialog box will then
appear at the far right end of the menu. Position the cursor over this box
and click, and a new window will appear as shown in Figure 10.24. Select
