1
$\begingroup$

I have a longitudinal data set and have fit a number of different regression models [ols, poisson, binomial....and more]. I want to justify the final model selection and assumed one method might be to use MAE and RMSE to show which 'fits' the data best. Is this approach one that has academic precedent? Is there some reason you should not use such to assess model fit?

$\endgroup$
1
  • 2
    $\begingroup$ That is the main approach that is used to select "better" models, of course assuming this is on a separate set. and not on the same training set (or similar alternatives). $\endgroup$ Jul 12, 2022 at 11:19

1 Answer 1

2
$\begingroup$

Many would opine that one should rather choose model form by knowledge of the phenomenon being studied, and the goal of the analysis. Some posts mostly arguing this way is

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.