I have a dataset in which I measure the response variable of a number of individual animals. I used this response variable as the dependent variable in a linear mixed model, for which I have 5 explanatory variables. Two of these explanatory variables are nested if they are entered in the same model, as they refer to the individual animal, and the research site they are situated on. So it would be site(individual). If I do not nest them but enter them separately, the program runs into calculation errors, and throws one of the two variables out. I use SPSS. I then created a whole number of different models, with all the different combinations of explanatory variables, and ranked them according to their AIC. I then found out that a number of models had exactly the same AIC. These were the models which, all else being equal, contained either just individual as an explanatory variable, or site(individual). If the model, all else being equal, only contained site, the AIC was different.
Now my questions are: a) do these results make sense? I cannot think of what I am doing wrong, but I am a bit worried about the exact same value for the AIC.. b) can I compare models this way - comparing nested variables with un-nested ones?
