I am working on a problem in which I have daily Sales and daily Offline and Online Marketing Spends. I built a linear regression model where Sales = f(Online Spend, Offline Spend, Lag Sales) at rolled-up weekly level. R square from the model is 92%; however, my intercept is negative, which does not make sense as even intercept should give the unexplained sales which should be non-zero in this case.

enter image description here

Am I missing something here? What could be possible reasons why the model is throwing a negative intercept? Any alternative suggestion/approach is much appreciated.

  • $\begingroup$ Numerous possibilities, including (1) linear form is a bad idea for your data (2) linear form is OK and your intercept corresponds to values a long way outside the range of your data and is nothing much to worry about. Just staring at numerical output can't give you enough assessment: you should be plotting your data and your results. A scatter plot matrix of your data and added-variable plots after the regression would help. $\endgroup$
    – Nick Cox
    Apr 4 '20 at 11:06
  • $\begingroup$ On "versus" see stats.stackexchange.com/questions/146533/… $\endgroup$
    – Nick Cox
    Apr 4 '20 at 11:07
  • $\begingroup$ Non-zero could mean negative. I think you mean positive. $\endgroup$
    – Nick Cox
    Apr 4 '20 at 11:09

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.