5
$\begingroup$

Could you please answer how should we approach this situation in linear regression problem: The target variable is distributed in a skewed fashion(50% of the values lie in the range 0-300 and 40% in 300-500 and 10% in remaining 500-1000. Will this cause a problem to my regression model and how will you approach to tackle with such problem?

$\endgroup$

1 Answer 1

4
$\begingroup$

As far as I know, you should look at whether or not your residuals are normal or not. If they are not normal, then you many consider transformation. For this, I would say you need to apply a log transformation, given it is fairly right skewed data. Applying a log transformation would help center the data into a more Gaussian distribution, where the data is more centrally located.

$\endgroup$
2
  • $\begingroup$ Could there be other potential alternative approaches if transformation is not preferrable? $\endgroup$
    – Krati A
    Aug 6, 2022 at 16:04
  • $\begingroup$ You can also try a generalized linear model if there is an appropriate link function that supports it. $\endgroup$ Aug 6, 2022 at 20:33

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.