These methods are well-understood
and most developed. Both models assume interval measurement and normally
distributed errors that are homogeneous across groups. The weak aspect of these
methods is that they only estimate and compare the group means and not
informative about individual growth. Furthermore, as an assumption, these
methods must have fixed time points. That is, each subject should have evenly or
unevenly spaced time points. The ANOVA model is given by: yij = μ + αi + βj+ εij
at i = 1… N, j = 1…n . where μ = grand mean, αi = individual difference
component for subject i (constant over time), βj = effect of time (same for all
subjects) and εij = error for subject i and time j. It is assumed that the random
components are distributed as αi ~N(0, σ2
α) where σ2
α is the between-subjects
variance, and εij~N(0,σ2
ε) , with σ2
ε is the within-subject variance. The
variance-covariance structure for yij is compound symmetry for ANOVA, which
assumes the covariate to be of the form σ2 + γδij for unknown parameters σ2 and γ
and where δij equals one for i = j and zero otherwise.