the insight that total variation can be split into "explained" (between-group) and "unexplained" (within-group) components. This idea extends naturally to regression (Ch. 22-23), where R-squared is the proportion of variation explained by the model.