Table 1 One-way ANOVA for data distributed according to the two-way mixed-effects model (1).
From: Design effect in multicenter studies: gain or loss of power?
Source | DF | SS | E(MS) |
|---|---|---|---|
Group | 2 - 1 |
|
|
Residual | N - 2 |
|
|
Total | N - 1 |
|
- When data are distributed according to the mixed model (1) but analyzed by performing a one-way ANOVA – as if data were distributed according to model (4) – the expectation of the residual mean squares (denoted
in the framework of model (4)) can actually be expressed as a function of 
and 
, the variance components associated to the true underlying statistical model (i.e. the mixed model (1)).
in the framework of model (4)) can actually be expressed as a function of 
and 
, the variance components associated to the true underlying statistical model (i.e. the mixed model (1)).