I am not sure what to do with this regression residuals.

Regression Diagnostics

The Normal Q-Q plot is indicating that residuals are non-normal (not too extreme) and residual variance is not constant (not too bad). My input variables are factor variables so I am not able to normalize it.

What do you guys think about this plot? Should I continue with this plot and say that assumptions are not violated and continue?

In fact, I think assumptions are violated slightly but not extremely. What should I do if I think assumptions are violated?

  • $\begingroup$ I didn't know about that!! great ! $\endgroup$ Jun 5 '13 at 12:33
  • 1
    $\begingroup$ It looks to me like you've got a few odd cases (they show in lower left in QQ plot). Those cases should be investigated. If they have errors, that may solve the problem. $\endgroup$
    – Peter Flom
    Jun 5 '13 at 12:48
  • 3
    $\begingroup$ You can't judge normality from the QQplot when the residuals exhibit curvature and the variance isn't constant; with those other assumptions violated the normality isn't readily assessed. $\endgroup$
    – Glen_b
    Jun 5 '13 at 13:40
  • $\begingroup$ data: residuals(fit) W = 0.9554, p-value = 0.000284 But I think for these kind of tests, It is unlikely to not reject the null(H0=normal) $\endgroup$ Jun 5 '13 at 14:13

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.