Statistical Procedures with the Law of Initial Values

Summary This paper reviews the literature on the Law of Initial Values (LIV) with respect to various definitions of the phenomenon and statistical methods for expressing and evaluating the LIV. An issue basic to each of these areas is whether or not and when to use difference scores. Most early statements and research articles accept the use of difference scores with little question, perhaps because of the phraseology in the original definition of the LIV. The present evaluation indicates that no additional information is gained by choosing difference scores over final scores and that difference scores involve additional interpretive burdens when evaluating treatment effects. Recommendations are made for the standardization of statistical indicators of the LIV. Finally, although conceptually more complex than some procedures that have been proposed, regression analysis appears to be the most appropriate statistical tool for dealing with the LIV.