I am working on a banking-related logistic regression problem. Several of my variables are whether someone has an account of some sort, and there is a subsequent variable with the balance of the account (or zero if they don't have an account).

By themselves, both variables are significantly related to the target variable; however, I worry including both will give me multicollinearity problems and probably other issues I am not aware of.

How would you recommend approaching determining which variable to keep?

  • $\begingroup$ Hi, have you estimated the model and checked for multicollinearity? $\endgroup$ – T.E.G. Aug 22 '20 at 13:04
  • $\begingroup$ We determined that some binary/continuous combos were redundant while others contribute unique information. For example, a binary variable for the presence or absence of a Certificate of Deposit and a continuous variable indicating the value of the certificate of deposit is redundant because you cannot have a CD with a zero dollar value. On the other hand, in the case of a binary and continuous checking account variable combo, both are valuable because it is possible to have a zero dollar (and even negative) checking account balance. $\endgroup$ – Eric Sims Aug 25 '20 at 23:01

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.