An epidemiological application of a Bayesian nonparametric smoother based on a GLMM with an autoregressive error component