I am working on determining how to handle some of my models. I found that the session with trap specific behavior model (g0~session+bk, sigma~session) ranks the highest for a number of groups.
However, when I run (g0~session, sigma~session), I get maximization errors. (Different errors for different groups.) One example is
Warning messages:
1: In secr.fit(M, model = list(g0 ~ session, sigma ~ session), :
probable maximization error: optim returned convergence 1. See ?optim
2: In secr.fit(M, model = list(g0 ~ session, sigma ~ session), :
at least one variance calculation failed
I do not find any glaring errors in the results of the more complex model (i.e. parameters appear to be estimable and within reasonable range). Is there any reason why the more complex model would work while the less complex does not work? Or should I assume that there is likely an error with the more complex model, even though it does not come up with warnings?
Is there a source someone might recommend to read about this issue further?
Thank you for your help!