In recent years, copula models have become a popular method for modeling correlated data, and have been widely applied in many field of studies. Although one can use the copula models to construct multivariate distribution easily, there is no research discussing the robustness of copula models so far. The purpose of this thesis is to investigate the robustness property of the copula models under model misspecifications. We also compare copula models with the multivariate negative binomial model.