ABSTRACT
This chapter presents a more detailed examination of residuals in multiple variable prediction equations. Several different types of residuals will be defined and discussed and particular attention will be given to how residuals can aid the data analyst in checking the error assumptions, determining model misspecification, and detecting extreme observations and outliers. The chapter contains definitions and discussions of the more commonly used residual types. It also presents some useful graphical techniques associated with residual analysis that aids the data analyst in correctly specifying a model. The majority of the graphical techniques and statistical measures employed to isolate outliers utilize studentized residuals since raw residuals as well as standardizedd residuals can fail to properly identify outliers due to variations in their standard errors. A relatively new graphical technique that plots residuals against deleted residuals can be extremely helpful in detecting outliers.
