Let’s put a summary of the common linear regression evaluation metrics. What they mean and how to use them.
Mean Absolute Error (MAE)
Meaning:
you should expect you predictions are off MAE from the true value.
pros:
cons:
- need to compare with other MAE to check if this MAE is good/bad
Root Mean Squared Error (RMSE)
Meaning:
squared root of MAE. Always greater than MAE
pros:
- More sensitive to outliers than MAE, so RMSE is most useful when large errors are particularly undesirable
- When used as a loss function, easilier to compute gradient.
cons:
- need to compare with other RMSE to check if this RMSE is good/bad
R-squared (R2)
Meaning:
R-squared = Explained variation / Total variation,
pros:
- A intuitive measurement of how well your predictions fit the observations.
- Don’t need to compare with other model, value tell you whether it’s good or bad
cons:
- R-squared cannot determine whether the coefficient estimates and predictions are biased.
- Adding a predictor will ALWAYS increase the R2 score
Adjusted R-squared (Adjusted R2)
Meaning:
Add penalty of number of predictors to R2, to solve the pitfall of R2