Title: Hierarchical Modeling - Some Environmental Applications

Abstract: 

The funny thing about hierarchical models is that, after you build one, you begin to see them everywhere. I'll describe how hierarchical models have been applied to microbial contaminants in drinking water. In most cases, the models perform very well, with high-level parameters describing how random effects vary across meaningful subsets of data. In one application, however, the multilevel model appears to explain significant differences between subsets, but a much simpler one-parameter model has significantly greater likelihood. How can that be?!? We'll address that and, if time allows, demonstrate how parameters are estimates using MCMC sampling in both R and OpenBUGS.