I have a dataset of 550 samples of Nests for which I know how many nestlings died or survived. I performed a binary proportion model: cbind(died,survived)~ variable 1+variable2+(1|Colony/Nest_ID),family=binomial) I would like to estimate the fit of the model.
But setting aside data for AUC curve seems impossible as many of the NestIDs in the random effect exist only once or twice and unavoidably always one will end up only in the testing dataset and will not be present in the training dataset. Are there any other methods of goodness of fit measures I could use?