You run a multiple linear regression predicting final exam score from (1) hours studied, (2) class attendance, and (3) test anxiety. In SPSS, the Normal P–P Plot of Regression Standardized Residuals shows strong S-shaped curvature, but the scatterplot of standardized residuals vs. standardized predicted values looks randomly scattered (no funnel shape). What is the best interpretation?
(Assume n = 120 and no extreme outliers.)
Linearity is violated, so you should immediately drop one predictor until the P–P plot looks straight.
Normality of residuals is questionable, but homoscedasticity looks okay; inference (p-values/CI) may be affected more than prediction.
Homoscedasticity is violated, so you should use a Mann–Whitney U test instead of regression.
Multicollinearity is definitely present because an S-shape indicates correlated predictors.