CompTIA DataX DY0-001 (V1) Practice Question

Adjusted R² adds a penalty for model complexity by multiplying the unexplained variance term by (n − 1)/(n − k − 1). The model with the highest adjusted R² therefore offers the best expected generalization while avoiding unnecessary predictors. In the table, M2 has the largest adjusted R² (0.809), so the 10-predictor model strikes the best balance between fit and parsimony. M3's raw R² rises slightly, but the heavier complexity penalty pushes its adjusted R² below both other models, signaling possible overfitting. M1 is simplest but gives up explanatory power compared with M2. The small numerical spread is precisely what adjusted R² is designed to evaluate; no additional test is required to see that M2 is preferred when this metric is the decision criterion.