When should I use estimated marginal means instead of observed means? Use EMMs whenever your design is unbalanced, whenever you have covariates (as in ANCOVA, where means are adjusted to a common covariate value), or whenever you want to make pairwise comparisons that respect the full model.
