Hello,

I have a randomized controlled trial with three measurement points (T1 = baseline, T2 = posttest, T3 = follow-up). Y is continuous and X is a 4-level factor. My research question is about the comparison between four treatment groups at posttest and then at the follow up. I'm not interested in between-groups difference of change over time, only whether the means at these two time points differ between groups, adjusted for their baseline scores. Therefore I conducted ANCOVA: separate analyses for outcome at Time 2 and Time 3 with T1 as covariate. Reviewers pointed out that this is not the best approach since data is longitudinal and Y at three time points is surely correlated. That's fair enough. They suggested GEE (generalized estimated equations), mixed linear models or just plain repeated measures ANOVA. However, as far as I understand, GEE and mixed models - while allow for controlling for dependency in observations - answer the question of TimexFactor which, as stated above, is not my research question. Is ANCOVA an appropriate solution? If not, how can I take the dependency of observations into account and at the same time answer my question? I would greatly appreciate your input.

Nina

More Ewelina Smoktunowicz's questions See All
Similar questions and discussions