1
$\begingroup$

RE: Model selection using misclassification rate in forward selection of logistic regression equation

A small misclassification error is good. Keep that factor in the model when doing logistic regression forward selection?

$\endgroup$
  • 1
    $\begingroup$ That is probably a very bad idea. Logistic regression is not a classifyer, misclassification is not a proper score function. Search this site for "model selection logistic regression". There are already many good posts! $\endgroup$ – kjetil b halvorsen Dec 12 '16 at 0:51
  • $\begingroup$ Your reply discussed a score function. I looked at the other posts briefly. Exactly what do you mean by a score function? Can you give a definition? $\endgroup$ – Mary A. Marion Dec 13 '16 at 5:35
  • $\begingroup$ en.wikipedia.org/wiki/Scoring_rule $\endgroup$ – kjetil b halvorsen Dec 13 '16 at 13:05
  • $\begingroup$ I found: The score function, 𝑢(𝜃), is the derivative of function 𝑙𝑜𝑔𝑓(𝑦|𝜃) with respect to the parameters. Thus it has to do with the maximum likelihood estimators. Misclassification error for logistic regression is thus not the best way to evaluate logistic regression models. My question then becomes is it necessary to use maximum likelihood for model comparisons? $\endgroup$ – Mary A. Marion Dec 13 '16 at 14:16
  • 1
    $\begingroup$ Done. I saw it earlier. I am now convinced. I ran 16 models using the misclassification error for model selection. I have been trying to be consistent in the use of statistics. The information about the error is still useful $\endgroup$ – Mary A. Marion Dec 14 '16 at 19:36

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.