How to compare models

The primary places you want to look when comparing models are...

  • Adjusted R2
  • Root Mean Square Error (RMSE)

The higher the Adjusted R2, the more variability that is explained by the model!

The lower the Root Mean Square Error (RMSE), the less error in predicting each of the observations exists!

Give me an example...

Scenario: Which model is better? Model 1 or 2? And why?

Model 1
Model 2

Model 1
Adjusted R2 = 0.78621
Root Mean Square Error (RMSE) = 43.8180

Model 2
Adjusted R2 = 0.86731
Root Mean Square Error (RMSE) = 34.5210

Adjusted R2
0.78621 (Model 1) < 0.86731 (Model 2)

RMSE
43.8180 (Model 1) > 34.5210 (Model 2)

Answer: Model 2 is better, as it has a higher Adjusted R2 and lower Root Mean Square Error (RMSE).

Activate AutoScroll