EXPLORING THE EFFECTS OF DISTORTING CLASSICAL LINEAR REGRESSION ASSUMPTIONS

Abstract : The cost analyst is often called upon to develop estimating relationships utilizing data on past experience. Even when the data base is adequate enough to permit regression analysis to be used, there is seldom enough data available to test whether or not the standard assumptions are met. The paper presents the results of a study in which the data problem was overcome by the use of Monte Carlo techniques. The results of several distortions of the standard case which appear likely to occur in practice were examined and compared to the standard case. The objective of the study was to demonstrate the effects of violating the standard assumptions in a very concrete manner. Many of the results could have been obtained in general mathematical form; however, the desire was to demonstrate the results empirically.