Is it possible for a feature not correlated (or faintly correlated) with a dependent variable to become important in a machine learning model?

  • $\begingroup$ It depends on the variable and the problem you are analyzing. Can you provide ad example dataset? $\endgroup$
    – Inuraghe
    Commented Mar 9, 2022 at 16:55
  • $\begingroup$ This is a general question, not specific to any data. Is it a possibility? $\endgroup$
    – NAS_2339
    Commented Mar 9, 2022 at 17:03
  • $\begingroup$ I'm looking for an answer with an example where non-correlated data become important feature in the model $\endgroup$
    – NAS_2339
    Commented Mar 9, 2022 at 17:19
  • $\begingroup$ In my example here, neither feature has any correlation with the color, yet the two features, jointly, almost perfectly predict the color. $\endgroup$
    – Dave
    Commented Aug 2, 2023 at 1:19

1 Answer 1


Suppressor variables are an example. Two good explanations are here and here but you can search and find a bunch more references.

A common view is a suppressor sharpens another variables relationship to the target. The suppressor is correlated with another feature and not the target but the correlation is in the area of the residuals of the other feature to the target.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.