More on Absolute Error, Etc.

Abstract

It is argued on statistical and conceptual grounds that absolute error can be a meaningful measure of performance. When approximately half of each group performs consistently on either side of the criterion, groups can have equal algebraic error and variable error means, leaving absolute error as the only dependent variable to discriminate between the groups. Additionally, absolute error is invariably the dependent variable subjects are asked to minimize in the experimental situation. If researchers persist in measuring motor performance through error scores, judicious use of absolute error is appropriate. Requests for reprints should be sent to K.M. Newell, Institute for Child Behavior and Development, University of Illinois, Champaign, IL 61820.