Specifying a random effect in GLIMMIX

[deleted_user]

Hi. I'm modeling a binomial response with a logit link and am wondering how best to model the unique covariance structure of two different study sites simultaneously (one has positive spatial autocorrelation in the residuals, one does not). Thoughts? I'm thinking using "_residuals_" in the RANDOM statement might be a key, but not sure the R matrix is what I should be focusing on.