Search

Linear regression model is commonly used to analyze data from many fields. Sometimes the data under research contains outliers, and it is important that these outliers be identified in the course of the correct statistical analysis. In this article we used genetic algorithm (GA) with three type of objective functions,Akaike information criterion (AIC), Bayesian information criterion (BIC), and Hannan–Quinn information criterion (HQIC) to detect the problem of masking and swamping outliers in linear regression model . Two well – known data sets have been studied and we conclude that GA doing-well in detection these type of outliers when using AIC and HQIC comparingwithBIC.

ABSTRACT In this paper we compute Akaike's Information Criteria (AIC), Bayesian Information Criteria (BIC) and Schwarz Bayesian Criteria (SBC) for all possible for independent variables. The three criteria are measurement of the difference between a given model and the real model, and the model with smallest (AIC), (BIC) and (SBC) compared with other models is the best one. This study is applied to data gathered from children infected with Thelasymia disease, and the data are analyzed by using SAS to evaluate which of the models is the best model to determine the best subset of variables that minimize the information criteria among all the variables in the study. We compare the three criteria with model diagnostic like root mean squared error (RMSE), Mallow's and adjusted .