Further comments: Misuse of analysis of covariance

Attempts to clarify some points of disagreement between D. A. Sprott and S. H. Evans and E. J. Anastasio . Examination of Sprott's discussion reveals that much of his argument rests on assuming an unconventional random effects model for the analysis of covariance. Sprott used only 1 regression term in his covariance model, which is tantamount to assuming homogeneity of within-between regression and inconsistent with the presence of a separate regression effect associated with treatments. It is reasserted after consideration of Sprott's comments that valid use of the analysis of covariance requires that the covariate be unaffected by the treatment. Consideration of a regression approach when the assumptions of the analysis of covariance are violated is recommended.