1
$\begingroup$

I have a multiple linear regression model with a square-root transformed dependent variable.

I want to report % deviance explained for each parameter. I assume I need to back-transform in some way, but I am not sure how. Or is it okay to report % deviance explained on the original (non-transformed) model?

I am using R to calculate model deviance, if it matters. I appreciate any insight on this issue.

$\endgroup$
  • $\begingroup$ You mention deviance. It sounds like you're fitting a GLM. Why transform? $\endgroup$ – Glen_b Jan 1 '15 at 4:00
1
$\begingroup$

The % deviance explained is calculated by comparing the predictions (fit$fitted.values in R) for Y to actual values of Y. It does not take into account any transformations on the explanatory variables. You might need to take into account the fact that you tested a raft of transformations when doing inference. With each transformation you are implicitly increasing the number of degrees of freedom for your numerator. R will report an adjusted R^2 but to be true to your modeling process you might need to recalculate it.

| cite | improve this answer | |
$\endgroup$

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.