Absolute Error

Abstract

School of Physical Education and Recreation The University of British Columbia The purpose of this investigation was to re-examine the measures of algebraic error (CE), absolute error (AE), and within-S variance (VE). in an attempt to arrive at unambiguous definitions regarding the statistical and logical meanings of these three performance variables. Under the assumption of a normal distribution, the statistic AE is completely dependent on CE and VE and thus can be predicted from them. This shows that all information contained in the term AE is contained in either CE (when the ratio $$\text{C}\text{E}/\sqrt{\text{V}\text{E}} > \text{2}\text{.0}$$), or in VE (when CE ≅ 0.0) or in a weighted combination of both CE and VE (when $$0<\text{C}\text{E}/\sqrt{\text{V}\text{E}}<2.0$$). Examples are provided to show the type of errors made in interpreting significant findings when AE is the dependent variable.