The statistical analysis of mean differences that are traced back to the effects of one or more factors (or independent variables). The simplest ANOVA is a one-way design in which N subjects are (randomly) allocated to a number of different levels of a single factor. Strictly speaking, ANOVA does not analyse variance per se, but rather sums of squares. Thus, the total variation in observations is partitioned into the between-groups sum of squares (between level means), and the within-groups (or residual) sum of squares (differences between subjects in the same group). The desired outcome in most cases is that the between-groups sum of squares is greater than that for the within groups. This is the F-ratio named by George W. Snedecor (1881-1974) after Ronald A. Fisher (1890-1962) who originally developed the test.
See Additive model, Balanced or orthogonal) design, Compound symmetry, Design matrix, Effect size, Error term, Error variance, Generalized linear model (GLM), Greenhouse-Geisser epsilon, Homoscedasticity, Mixed-effects models, Multivariate analysis of variance (MANOVA), Orthogonality, Planned comparison, Polynomial analysis of variance, Repeated measures analysis of variance, Residuals, Sphericity, Sphericity assumption