👉 For ANOVA, both the homogeneity of variances and the normality assumptions concern the errors of the model , so they should be assessed on the residuals . Below is the precise reasoning, with practical nuances. 1. What ANOVA actually assumes The classical ANOVA model is: Y i j = μ + α i + ε i j Y ij = μ + α i + ε ij with the assumptions: Normality : ε i j ∼ N ( 0 , σ 2 ) ε ij ∼ N ( 0 , σ 2 ) Homoscedasticity : V a r ( ε i j ) = σ 2 Var ( ε ij ) = σ 2 for all groups Independence of ε i j ε ij So both assumptions apply to the errors , not to the raw response Y Y . 2. Consequences for diagnostics ✅ Normality Should be assessed on residuals , not on original data. Raw data can be non-normal simply because group means differ. Correct tools: Q–Q plot of residuals Histogram of residuals Shapiro–Wilk test on residuals (with caution) ✅ Homogeneity of variances Also concerns residual variance ...