Information Technology Reference
In-Depth Information
Between Subjects
Variance
Figure 10.1
Total variance of the depen-
dent variable partitioned into between-
subjects and within-subjects variance.
Within Subjects
Variance
portion of the variance of the dependent variable can be attributed to
differences between the subjects.
The between-subjects variance in a within-subjects design captures the
extent to which
individual differences
were observed between the subjects
in the study. Once the individual differences are accounted for (once the
variance due to between-subjects variance is identified) in the ANOVA,
then the remaining variance is within-subjects variance. This is illustrated
in Figure 10.1. The circle represents the total variance of the dependent
variable. We have partitioned the total variance into two general regions:
between-subjects variance and within-subjects variance.
10.5.3 WITHIN-SUBJECTS VARIANCE IN A WITHIN-SUBJECTS DESIGN
A within-subjects design requires all of the participants to be measured
under each study condition. With the very same participants represented
at each level of the independent variable, any mean differences that we
observe between the conditions cannot be attributed to differences in the
participants experiencing those conditions (i.e., all conditions are expe-
rienced by exactly the same people). Thus, in comparing one level of the
independent variable to its other levels, there are no between-subjects
differences (there is no between-subjects variance) with which to con-
tend. Because the same participants are involved in such comparisons of
means, the variance represented by the means of the different conditions
represents the within-subjects variance. In that sense, it is often said that
in within-subjects designs, participants serve as their own controls.
In this
way, within-subjects designs are very powerful. They statistically remove
(account for) a potentially major source of error, namely, individual dif-
ferences among the participants.
If within-subjects designs are so powerful, why are they not used all or
most of the time? The answer is that many of the independent variables we
study are not amenable to a within-subjects manipulation because of the
adverse carry-over effects that would be involved. That is, for many of the
constructs we study, having exposed people to one condition (one level
of the independent variable) sufficiently changes them so that they are no
longer naıve with respect to any other research condition. Furthermore,
