0
$\begingroup$

I obtained the following residual versus fits plot using R:

enter image description here

Does this plot obey the simple linear regression assumption of constant variance? I think it doesn't since the residuals seem to get more dispersed as the 'fits' values increase, but I am hoping that I can get a second opinion on this somewhat ambiguous graph. Are there any additional graphs that I can draw in R to determine whether the residuals obey constant variance?

$\endgroup$
1
  • $\begingroup$ Seems like your line at about 20 degrees up from horizontal is a more pressing issue ... $\endgroup$ Sep 4 '20 at 21:24
0
$\begingroup$

These things are sometimes subjective if its borderline. It could be better but conversely you only get perfect clouds of data in textbook examples.

If possible try and identify the points in the line with a shallow slope that increases with fit, the rightmost of which is the highest residual at approximately at $3.5x10*5$

You could plot studentized residuals, and this removes the unequal variances of the residuals caused purely by their position within the regressors.

Wikipedia has a page on heteroskedasticity, on which there is a section on tests in regression. The Breusch–Pagan test is quite straightforward.

$\endgroup$

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.