I am trying to remember what is the name of the effect when feature has a high weight in a linear regression, which logically should not have? For example if we train and test model to predict stock price and we have a moon phase as feature and after training it has high weight. We understand that moon phase has nothing to do with stock prices but model says that it has high weight. What is this effect name and how we mathematically find those features? I just forgot the name of that test/effect)))

  • $\begingroup$ I don't know but this could be due to noise in the data, otherwise I would just call this a counter-intuitive result (e.g. more people than you think might play the stock market based on astrology predictions). $\endgroup$
    – Erwan
    Sep 1, 2022 at 9:44
  • $\begingroup$ Do you maybe mean spurious correlation? $\endgroup$
    – buddemat
    Sep 2, 2022 at 18:26


Your Answer

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

Browse other questions tagged or ask your own question.