The Utility of Within-Subjects Variables: Estimates of Strength

Abstract

Procedures to estimate the magnitude of a treatment effect are examined for analysis of variance designs which have repeated-measures. It is shown that the estimate of the magnitude of an effect logically follows from the expected mean squares associated with the design in question. In particular, it is argued that the estimate of utility (strength of effect) for a within-subjects variable should be adjusted to reflect the underlying ANOVA model; viz., the inter-subject variability should be removed from the utility estimate.