0
$\begingroup$

I have carried out a model selection process and averaged the models within the top 2 AICcs (using MuMIn in R). I would like to test the model fit using both the chisq and validation/residual plots. I'm unsure how to write the code for this, or indeed if it is even possible?

Any help would be appreciated.

Improve this question
$\endgroup$
3
  • $\begingroup$ Trying to understand at which step you are having a difficulty... Are you able to do that for other models rather than what MuMIn returns (so that it is MuMin that is the trouble)? Or are you unable to do that for any model (so that MuMIn itself is not the trouble)? What do you mean by test the model fit using ... chisq? $\endgroup$
    – Richard Hardy
    Mar 19, 2015 at 11:00
  • $\begingroup$ I can do both diagnostics for singular models using MuMIn and other packages. The issue is how do you run diagnostics on averaged models. $\endgroup$
    – user13641
    Mar 19, 2015 at 13:32
  • $\begingroup$ How is an averaged model different from a simple model so that these tests cannot be readily conducted on it? Is it due to MuMIn model output belonging to some "weird" class in R, or is there also a conceptual difference? If it is just a "wrong class" problem, have you tried changing the class of the model to one that would fit your needs? $\endgroup$
    – Richard Hardy
    Mar 19, 2015 at 14:08

1 Answer 1

Reset to default
1
$\begingroup$

The issue is that an "averaged" model through MuMIn is actually stored as a list of the input models with a few new attributes. It is not actually the "output" model. Every time you print or summarize the model.avg output, it re-generates the averaged outputs from the list for display. You will need to manually generate the components of the diagnostic plots using predict and then do the math.

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.