Hello Stack Exchange community,

I'm running a regression on a survey dataset which consists of 45,381 individuals in 36 countries using Stata. My model is specified as:

reg income religious `controls' [pweight=Combinwt], cluster(village)

where Combinwt is the probability weight and standard errors are clustered at the village level, since the survey is performed by randomly choosing villages in a region and then randomly choosing individuals within that village.

When I ask Stata to produce a residuals vs. fitted plot, this is the result:

Residuals vs. fitted plot of OLS regression I have never seen an rvf plot that looks like this before. Does anyone have any ideas for what could explain the shape of the plot?

Best regards,


If you watch closely, you will find out that the slope of the lines is -1. I.e. the values of y along every one of the lines are exactly equal. That is you get this lines only because you have great many cases for which the left hand side variable is equal. The lines are visible because relative to the number of observations there are not that many different values of y.

  • $\begingroup$ Thank you for a really clear answer! It's true, I only have about 40 distinct values for my left hand side variable. Follow up question: Would you say that this pattern is problematic for statistical inference? $\endgroup$ – Tómas Bjarnason Mar 1 '20 at 21:52
  • $\begingroup$ @TómasBjarnason Well, not necessarily. Of course if you are modelling income and have many zeros on the left hand side you may want to consider a Tobit model. $\endgroup$ – Grada Gukovic Mar 1 '20 at 21:58

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.