0
$\begingroup$

In a regression context, specification of the regression equation is rather important. When we add predictors to the regression specification which are not relevant can cause a misspecification issue, and therefore hurt the model's accuracy.

I was wondering, whether in the random forest framework this is a big issue? I have many predictors in my dataset, should i include all of them. If no, what is the reason? And how to select which predictors should be included? Thank you!

Improve this question
$\endgroup$
4
  • $\begingroup$ I've partly addressed this in this answer: stats.stackexchange.com/a/288762/121522 $\endgroup$
    – mkt
    Commented Aug 5 at 8:15
  • $\begingroup$ As for how to select predictors, subject expertise is always helpful. $\endgroup$
    – mkt
    Commented Aug 5 at 8:16
  • $\begingroup$ @mkt This question is not a duplicate and the answer therein is not satisfactory. $\endgroup$
    – Sane
    Commented Aug 8 at 12:22
  • $\begingroup$ Feel free to edit your question to clarify the part that is still unclear or is distinct from the other thread. Note that it requires 3 members of the community to close a question as duplicate, so it's not just me who thought it was covered. $\endgroup$
    – mkt
    Commented Aug 8 at 13:24

0

Reset to default

Browse other questions tagged or ask your own question.