I like a mixed-effects model here, where you have subject as a random effect. This allows you to model all four baseline/follow-up points. You would then essentially test for differences in slope over the four points in time between your fixed treatment effect. An elegant solution to repeated measures analysis. When the model is balanced, the same could be achieved with MANOVA, but personally I favor the language of mixed models.
Brett . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
