Detection Of Outliers In The Linear Regression Model With RJMCMC Method

The research of detection of outliers in the linear regression model has been a hot topic all the time for the complexity of the real data. With RJMCMC method, the determination of the number of outliers in linear regression model becomes not hard as before.Firstly, we introduce the concept, cause and value of doing research in detecting outliers. Secondly, we show the equivalence theorems between mean shift model and outlier detection model based on the condition that errors in regression model are normally distributed. Thirdly, after RJMCMC method proposed by Sylvia Richardson and Peter J. Green is introduced in detail, this article shows theories and procedures about the application of RJMCMC method in linear regression model. Finally, this method is shown feasible and effective in outlier detection by being applied to a benchmark outlier dataset which is generated by Hawkins, Bradu and Kass. Besides, this article also proposes an effective method along with its detailed steps to test outliers.