From the column of Fin the summary table of Table 14.7c, we see that the main effect of
Sex is significant, as is the Time 3 Group interaction. Both of these results are meaningful.
As you will recall, the dependent variable is a measure of the frequency of use of condoms
(log(freq 1 1)). Examination of the means reveals adolescent girls report a lower frequency
of use than adolescent boys. That could mean either that they have a lower frequency of in-
tercourse, or that they use condoms a lower percentage of the time. Supplementary data
supplied by St. Lawrence et al. show that females do report using condoms a lower percent-
age of the time than males, but not enough to account for the difference that we see here.
Apparently what we are seeing is a reflection of the reported frequency of intercourse.
The most important result in this summary table is the Time 3 Group interaction. This
is precisely what we would be looking for. We don’t really care about a Group effect, be-
cause we would like the groups to be equal at pretest, and that equality would dilute any
overall group difference. Nor do we particularly care about a main effect of Time, because
we expect the Control group not to show appreciable change over time, and that would
dilute any Time effect. What we really want to see is that the BST group increases their use
over time, whereas the EC group remains constant. That is an interaction, and that is what
we found.Simple Effects for Complex Repeated-Measures Designs
In the previous example we saw that tests on within-subjects effects were occasionally dis-
rupted by violations of the sphericity assumption, and we took steps to work around this
problem. We will have much the same problem with this example.
The cell means plotted in Figure 14.3 reveal the way in which frequency of condom
use changes over time for the two treatment conditions and for males and females sepa-
rately. It is clear from this figure that the data do not tell a simple story.
We are again going to have to distinguish between simple effects on between-subject
factors and simple effects on within-subject factors. We will start with between-subject
simple effects. We have three different between-subjects simple effects that we could
examine—namely: the simple main effects of Condition and Sex at each Time, and the
Sex 3 Condition simple interaction effect at each Time. For example, we might wish to
check that the two Conditions (BST and EC) do not differ at pretest. Again, we might also
want to test that they do differ at FU6 and/or at FU12. Here we are really dissecting the
Condition 3 Time interaction effect, which we know from Table 14.7 to be significant.486 Chapter 14 Repeated-Measures Designs
Table 14.8 Expected mean squares with A, B, and Cfixed
Source df SS
Between subjects abn 2 1
Aa 2 1
Bb 2 1
AB (a 2 1)(b 2 1)
Ss w/in groups ab(n 2 1)
Within subjects abn(c 2 1)
Cc 2 1
AC (a 2 1)(c 2 1)
BC (b 2 1)(c 2 1)
ABC (a 2 1) (b 2 1)(c 2 1)
C 3 Ss w/in groups ab(n 2 1)(c 2 1)
Total N 2 1s^2 e1s^2 gps^2 e1s^2 gp 1 nu^2 abgs^2 e1s^2 gp 1 nau^2 bgs^2 e1s^2 gp 1 nbu^2 ags^2 e1s^2 gp 1 nabu^2 gs^2 e 1 cs^2 ps^2 e 1 cs^2 p 1 ncu^2 abs^2 e 1 cs^2 p 1 nacu^2 bs^2 e 1 cs^2 p 1 nbcu^2 a