A residuals-versus-fitted plot checks whether a regression or ANOVA model leaves systematic errors behind. It places residuals on the y-axis and fitted values on the x-axis. Residuals measure the vertical distance between observed and fitted values. section 5.7.5 section 14.5.2
Look for points hovering around the horizontal line without an obvious pattern. Curves or other structure can signal nonlinearity; changing spread can signal unequal variances. In an ANOVA, positive residuals mark teams above their category mean, while negative residuals mark teams below it. section 5.7.5 section 14.5.2
Inspecting an NBA ANOVA model
Fit an ANOVA model for standardized regular-season wins, grouped by points category, then request the residuals-versus-fitted and Normal QQ plots with plot(fit1, which = c(2, 1)). The residuals were less dispersed at the highest fitted values, suggesting that the equal-variances assumption was probably violated, though not substantially. section 14.5.2
plot(fit1, which = c(2, 1))
Run this R command to display the Normal QQ and Residuals vs. Fitted diagnostics for the ANOVA model. section 14.5.2
