I have a set of 9 models that I am running using R package MixSIAR to investigate the diet of Bottlenose dolphins in an estuary. The models have combinations of factors with random (social cluster, individual) and fixed (age, sex) factors. Two of the models have a nested structure (individual nested within a social cluster, sex/age nested within a social cluster).
For relative model fit I am using LOOic, with the model comparisons run through the MixSIAR package.
The model with the most weight has a negative LOOic value and I wanted to double-check that this is okay and the model indeed provides the best relative fit to the data?
I am also confused as the best model according to LOOic (although negative LOOic value) has the largest DIC value out of the 9 models. Any suggestions on what could be going on there?
I appreciate any guidance on this.
