What's the motivation of data transformation (like Box-Cox) to skew data?

Suppose I am only interested in predictive power and non-linear model. Does it help to do data transformation (like Box-Cox) to skew data?

  • $\begingroup$ I think it depends on the choice of model. If it's machine learning model, we have less assumptions so transformation is not that necessary. If it's a statistical model, there are many assumptions (like normality of data) to be met. $\endgroup$
    – John
    Mar 28 '16 at 6:08
  • 2
    $\begingroup$ It would be extremely unusual for anyone to use a Box-Cox transformation, or any transformation for that matter, purposely to make data more skew! $\endgroup$
    – whuber
    Mar 28 '16 at 19:03
  • 3
    $\begingroup$ @whuber I think this is an English issue. Perhaps the OP means apply this to data that are skew (using skew as an adjective) rather than to make them more skew (using skew as a verb). $\endgroup$
    – Peter Flom
    Jul 5 '18 at 13:41
  • $\begingroup$ Does this answer your question? How is the Box-Cox transformation valid? $\endgroup$ Feb 20 '20 at 18:16

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.