Multiply imputed synthetic data: evaluation of Hierarchical Bayesian imputation models