By definition, for a linear model we have that $$\text{AIC}=n\log(\frac{\text{RSS}}{n})+2p$$ where $n$ is the number of observations, $p$- the number of parameters and RSS - the residual sum of squares

My question is: If we have the Residual Deviance for a given GLM can we find the AIC?


  • 2
    $\begingroup$ This only works under the assumption of normality of errors. Otherwise you would have likelihood in place of RSS. $\endgroup$ Apr 11, 2018 at 9:59
  • $\begingroup$ The definition for AIC is well defined on Wikipedia. However, how the formula looks like depend on what GLM family you're talking about. $\endgroup$
    – SmallChess
    Apr 11, 2018 at 9:59
  • $\begingroup$ My issue (poorly written in the question) is if I'm given only the Residual Deviance, can I find the AIC? Will edit the question $\endgroup$
    – asdf
    Apr 11, 2018 at 10:20
  • $\begingroup$ AIC is in general calculated as $$\text{AIC} =2k-2\ln(\hat{L}),$$ where $k$ are the number of parameters and $\hat{L}$ is the maximized likelihood of the model with $k$ parameters. $\endgroup$
    – BloXX
    Apr 11, 2018 at 10:33
  • $\begingroup$ But does this have any relation to the Residual Deviance? $\endgroup$
    – asdf
    Apr 11, 2018 at 14:47


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.