The aim of this thesis is to integrate three areas of statistical research---multilevel modeling, Markov chain Monte Carlo (MCMC) methods, and measurement error. Three distinct types of multilevel models are considered: random-intercepts models, random-slopes models, and models with complex variation at level-1. These models are fitted using MCMC and maximum likelihood methods and the fits are compared. Finally, the effects of measurement error in predictors are assessed for different reliabilities and adjusted for using MCMC methods. The results indicate that MCMC samplers with non-informative priors produce similar results to maximum likelihood estimates and adjust for measurement error in predictors effectively. In general, MCMC methods give smaller standard errors, making inferential statements more powerful, and facilitate the use of additional information to guide the measurement and regression process. The simulations were performed using S-plus and the multilevel model problems were formulated and solved using MLwiN.