I performed a multiple linear regression and after checking my results everything seems to be great, but when I checked the residuals I noticed this patterns in the attached figure and I am not sure if it's right or not. I think there are two (or maybe three) groups. Am I right? Or this is a a random distribution? If there are two or more groups, what should I do next?



  • $\begingroup$ well, If I were you I would check distribution of residuals since in MLR the residuals are supposed to be normally distributed. It should be available in Minitab as far as I remember. $\endgroup$ – theGD Oct 13 '17 at 0:12
  • $\begingroup$ The residuals are normally distributed (without outlier) $\endgroup$ – Billo Oct 13 '17 at 0:45

You seem to have 3 groups of predicted values. This looks like you have one categorical predictor that takes 3 levels and one or more continuous predictor, but that the categorical predictor has a stronger relationship. That is not, in itself, a problem. But if the residuals are related to the fitted values, that is a problem. However, you have so little data that it is hard to tell.

The fact that you have so little data makes me worry about overfitting.

  • $\begingroup$ How can I check of the residuals are related to the fitted values? $\endgroup$ – Billo Oct 13 '17 at 3:18
  • $\begingroup$ Ordinarily, you would look at a plot like this one. It's just that with so few subjects, it's hard to see any pattern. $\endgroup$ – Peter Flom Oct 13 '17 at 11:19
  • $\begingroup$ The usual way to express thanks here is to upvote. And, if your question is now answered, you can accept my answer. $\endgroup$ – Peter Flom Oct 15 '17 at 21:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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