Could you give an example of the situation when if you input wrong (economic-wise, but not mathematically) parameters the model could fit/forecast better?

  • $\begingroup$ You are not asking about spurious regressions, are you? tylervigen.com/spurious-correlations They fit well but have 0 forecasting power. Do you mean things that are not "wrong" but unexpected? Like if male underwear sales rise it signals the end of an economic depression? $\endgroup$ – Giskard Sep 21 '15 at 15:18
  • $\begingroup$ @denesp no. What I'm trying to find out is the possibility of for example we have Y=Model(X), and X is determined by some economics conditions which are not necessary for the Model to be solved/worked with. Then by disregarding those 'extra' economics conditions, and focusing only on the solely mathematical ones, we could find a better fit for the observed data Y. $\endgroup$ – An old man in the sea. Sep 21 '15 at 22:27
  • $\begingroup$ I guess this has to depend on how you define better fit. (Like $R^2$ or adjusted $R^2$.) It seems to me you can never get a better model by throwing away information/data. $\endgroup$ – Giskard Sep 21 '15 at 22:58
  • 2
    $\begingroup$ @denesp en.wikipedia.org/wiki/Overfitting $\endgroup$ – An old man in the sea. Sep 22 '15 at 8:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.