ABSTRACT
In clinical research, a parallel-group design with multiple assessments at a number of prespeciŒed time points post treatment is usually employed to compare treatment difference between a test compound and a control (e.g., a placebo control, a standard therapy, or an active control). Under such a study design, a one-way (treatment) analysis of variance (ANOVA) with repeated measures is a valid statistical method for assessment of treatment difference. In practice, however, it is not uncommon that a two-way ANOVA is wrongly used to assess treatment difference by treating the prespeciŒed time points as a class variable rather than a covariate from the same subject. It, however, should be noted that one of the primary assumptions for a two-way (treatment by time) ANOVA is independence among observations observed at different time points. As a result, the use of a two-way ANOVA is inappropriate since observations observed at different time points from the same subject are correlated. Thus, it is of interest to evaluate the validity of statistical inference obtained from the two-way ANOVA under the one-way ANOVA model with repeated measures.
