Testing the normality of errors in regression models with a forward approach

This paper presents a new robust method for testing the normality of errors in classical linear model. It gives information about both the true distribution of the errors of most of the data and the percentage of contaminated data.