Remarks On The Analysis Of Covariance In Repeated Measures Designs

The analysis of data collected in a design involving one between groups and one within subjects factor, each with two levels, and one observation per subject on the covariate, is considered. Conditions necessary for ANCOVA adjusted main effects and interactions to be meaningful are presented. Under these conditions the adjusted effects and ANOVA unadjusted effects are shown to be equal, and the variances of the unadjusted and adjusted effect are compared. The adjusted within subjects effect cannot be less variable than the unadjusted effect, while the adjusted between groups effect and interaction can be less variable than the corresponding unadjusted effect.