I am only presenting the most important parts of the printout, but you can see the rest by
running the analysis yourself. (The data are available on the book’s Web site as
WickMiss.dat.)502 Chapter 14 Repeated-Measures Designs
Information Criteriaa
–2 Restricted Log Likelihood 905.398
Akaike’s Information Criterion (AIC) 909.398
Hurvich and Tsai’s Criterion (AICC) 909. 555
Bozdogan’s Criterion 916.136
Schwarz’s Bayesian Criterion (BIC) 914.136
The information criteria are displayed in smaller-is-better forms.
aDependent Variable: dv
Fixed Effects
Type III Tests of Fixed Effectsa
Source Numerator df Denominator df F Sig.
Intercept 1 22.327 269.632 .000
Group 1 22.327 16.524 .001
Time 3 58.646 32.453 .000
Group * Time 3 58.646 6.089 .001
aDependent Variable: dv
Covariance Parameters
Estimates of Covariance Parametersa
Parameter Estimate Std. Error
Repeated Measures CS diagonal offset 2954.544 551.1034
CS covariance 2558.656 1026.581
aDependent Variable: dvI will not discuss the section labeled “Information criteria” here, but will come back to
it when we compare the fit of different models. The fixed effects part of the table looks just
like one that you would see in most analyses of variance except that it does not include
sums of squares and mean squares. That is because of the way that maximum likelihood
solutions go about solving the problem. In some software it is possible to force them into
the printout. Notice the test on the Intercept. That is simply a test that the grand mean is 0,
and is of no interest to us. The other three effects are all significant. We don’t really care
very much about the two main effects. The groups started off equal on pre-test, and those
null differences would influence any overall main effect of groups. Similarly, we don’t care
a great deal about the Time effect because we expect different behavior from the two
groups. What we do care about, however, is the interaction. This tells us that the two groups
perform differently over Time, which is what we hoped to see. You can see this effect in
Figure 14.5.
There are two additional results in the printout that need to be considered. The section
headed “Covariance Parameters” is the random part of the model. The term labeled “CS
diagonal offset” represents the residual variance and, with balanced designs, would be the
error term for the within-subject tests. The term labeled “CS covariance” is the variance of
the intercepts, meaning that if you plot the dependent variable against time for each sub-
ject, the differences in intercepts of those lines would represent differences due to subjects