In this paper , we ?rstly discuss the interaction between model selection and outlieridenti?cation based on Bayesian method, suggest that model selection and outlier iden-ti?cation should be identi?ed simultaneously, and the Gibbs sampling iterations basingon full conditional can re?ect this relation, so we choose this algorithm. Secondly, dataaugmentation is introduced to identify the model and outliers simultaneously basedon the posterior of model and outlier. Finally, for binary and ordinal data , we ex-press the process of simultaneously model selection and outlier identi?cation accordingGibbs sampling in detail and give two simulated examples to express my method andits e?ciency. |