Closed djhocking closed 10 years ago
I added site nested within HUC8 to match the formulation in the jpeg above. The only difference is that their are also year effects (crossed, not nested) and the coefficients with HUC8 are drawn from a multivariate normal distribution so that they can covary and we can estimate that correlation (rho). It was done as suggested in Gelman and Hill.
Add another level to the model so that the random site effect is nested within the random HUC. There may be challenges associated with this such as: