I am trying to determine model significance using:
1-pchisq(null deviance-residual deviance, null df- residual df)
I have 5 models:
- Four models were estimated with GLMs, which gave me null and residual DFs in the summary.
- The fifth model was estimated with the
lmer
function because of the nested structure of my data. This does not give me degrees of freedom for null or residuals.
How do I get my null and residual deviances and degrees of freedoms?
Thanks!
Best Answer
The concept of numbers of parameters and hence df in the lmer model is kind of fuzzy. Don't bother with it, and use AICc; you stand on firmer theoretical ground: http://warnercnr.colostate.edu/~anderson/PDF_files/TESTING.pdf