Estimation for an improved multilevel model based on MCMC algorithm

Independence among groups is assumed in traditional multilevel models. There is often spatial interaction between districts when data is grouped by geographical units. The individual will be influenced by adjacent regions, and assumption of level-2 residual's distribution in traditional multilevel model will be violated. Spatial statistical models are introduced into the multilevel model in order to deal with such spatial multilevel data. And Bayesian inferences based on MCMC method for fixed effects, variance-covariance components and spatial regression parameters in improved multilevel model are given.