I am performing a simple linear regression where y = income and x = education. I was just wondering if there exists p-value associated with the residual e in linear regression model to test its statistical significance. How does one obtain the p-value?

Thank you very much.

  • $\begingroup$ Why do you think you want such a $p$-value? $\endgroup$ Commented Jul 19, 2018 at 19:28
  • $\begingroup$ I want such a p-value because I want to find some portion of income that's not explained by education. $\endgroup$ Commented Jul 19, 2018 at 19:29
  • $\begingroup$ Interesting. I suggest thinking about what assumptions must be in place in order for any p-value to be calculated...and then whether such assumptions could be made to apply to your situation. $\endgroup$
    – rolando2
    Commented Jul 19, 2018 at 19:31
  • $\begingroup$ In that case you do not want those $p$-values. Instead you are looking for $R^2$: this is the proportion of the variance in income explained by education, so $1-R^2$ is the proportion of the varinace in income not explained by education. $\endgroup$ Commented Jul 19, 2018 at 19:31
  • 1
    $\begingroup$ (Note that in linear regression, the mean of residuals always takes the same value.) If you want to demonstrate that certain residuals exceed the mean residual by >abs(1.96 standard deviations), that should be easy to do. Given N observations, you have N opportunities for that to occur, and in a large data set you'll routinely have residuals out at 3 and 3.5 standard deviations. It's hard to see how these would yield a meaningful p-value tailored to any question. OTOH, you may find it profitable to look into influential points for what they do signify. $\endgroup$
    – rolando2
    Commented Jul 19, 2018 at 19:52


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.