Statistical Treatment of the Law of Initial Values (LIV) in Autonomic Research: A Review and Recommendation

Abstract

Three major techniques for statistically handling LIV are reviewed: (1) regression analysis as proposed by Laceyl2 (ALS); (2) the difference between standardized pre- and poststimulus scores as proposed by Heath and Oken, 11, 17 and (3) the intuitively appealing per cent change transformations. The proposed criterion of a satisfactory score is: The correlation between the initial level and a properly corrected score should be zero, and the adjusted score should have face validity. Mathematical analysis demonstrates that only scores built on the regression models will be guaranteed to have no correlation with initial level. Consequently, for comparing differences between groups'' response to experimental stimulation, covariance analysis of the difference scores (D) adjusted for initial level (X) is recommended. For comparing differences between individuals'' response to experimental stimulation, a difference score from the Covariance XD model, d [long dash] bDXx is recommended. In standardized form, this is equivalent to Lacey''s ALS.