Thanks for your reply!

My concern about whether HLM is appropriate is that the first tier (crime events) follow Negative Binomial Regression. So I think linear model may not be a good model for this project (I could be wrong).  That's why I am also thinking about Bayesian Hierarchical Spatial Model. However, bayesian models are always too complicated.

So, again, which model is better for this project:  HLM with Spatial, or GLIMMIX, or Bayesian Hierarchical Spatial Model?

Also, could you provide some links about the winbugs/R code for the related models?

Thanks!