I want to make a statement about model prediction qualities, comparing several models based on MAE and MSE. Considering the following results:

Model comparison

Is it valid to say considering models 1 and 2:

  • Model 1 outperforms two showing an improvement in MAE of 24% ((0.0792-0.1042)/0.1042).
  • Model 1 has significantly fewer large errors due to much larger relative difference in MSE of 63% ((0.0188-0.0502)/0.0502)?

Should the relative difference in MAE and MSE between any two models be identical if they perform idnetical on "normal data" as well as considering outliers?

In general I am wondering on if and how to compare these two metrics with each other. Is it possible to make an equal statement regarding R2, saying that although model 1 and 2 are not far apart by MAE, model 1 explains much more variance in the data?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.