It is well known that inferences based on least squares regression can be strongly influenced by only a few observations in the data, and the fitted model may reflect unusual features of those observations rather than the overall relationship between the variables, [6].