The expected mean squares (EMS) for the effects in a two-way ANOVA model are derived when sampling from a
finite population, for factors A and B. The A-effects and B-effects are represented by i and
j , respectively, in the model.
Model effects are studied for the random and mixed effects models. Thus, in terms of hypothesis testing, we are interested in the
expected mean square formulas when the random effects are sampled from finite populations. For balanced data, the EMS for the
A, B and AB interaction effects when the effects are sampled from a finite population are the same as the EMS for an infinite
population. For unbalanced data, the EMS when the effects are sampled from a finite population in factors A, B and the AB
interaction, the EMS are not the same as for an infinite population because the values that multiply the variance components
differ.