Residuals vs. Fitted Values plot

enter image description here

Can someone tell me which OLS assumptions might be violated according to the figure and why? Moreover, what are the consequences for the point estimators and the standard errors?

  • 1
    $\begingroup$ The first question is answered by stats.stackexchange.com/questions/76226/… $\endgroup$ – xan Feb 25 '17 at 23:19
  • $\begingroup$ is it correct to conclude that the model is nonlinear and that there is heteroskedasticity? $\endgroup$ – Moody707 Feb 26 '17 at 16:40
  • 1
    $\begingroup$ Possible important violations: (i) a covariate is missing, (ii) you used a linear model for a nonlinear relationship, or (iii) the data is not independent (which is related to (i), i.e., the missing covariate could be time or spatial coordinates). If any of these apply and are as strong as in this example, other violations such as heteroskedasticity are less important. $\endgroup$ – Roland Feb 27 '17 at 7:19

Browse other questions tagged or ask your own question.