I'm conducting regression analysis and I'm wondering if I should handle multicollinearity before or after my chosen feature selection method. The data I am using produces VIF values greater than 10 for 3 of the explanatory variables, so should I handle for multicollinearity by removing the inflation variables first so that my VIF values are all less than 10 or should I run my feature selection method first (Stepwise Selection in this case) and handle for multicollinearity after?

I've tried conducting feature selection both before and after adjusting for multicollinearity and each option gave me a different best model. However the latter still provided a model with higher multicollinearity and adjusting for it gave me less complex model than the former.



Your Answer

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