This paper discusses the type of dependence induced by the Generalized Additive Mixed Model (GAMM) approach to regression analysis with correlated data. In this framework, random effects are added on the same scale as the fixed effects. Dependence between out- comes is thus generated by their sharing of common/correlated latent variables. In many cases, this results in strong positive association.
Brouhns, N., & Denuit, M. (2002). Actuarial modelling of longitudinal claims data through GAMM’s : some methodological results (STAT Discussion Paper 0220). https://hdl.handle.net/2078.5/23638