I have a "3 level" linear growth model with time points nested in individuals and individuals nested in schools. Based on the nature of the sample and the research question(s) I know that regression to the mean will be an issue in these data potentially for individuals and schools. My question is this: What are different ways to "control" for regression to the mean in this case? I know that in OLS covarying pretest is an option. I have experimented with this in my current model but am not clear if I should include Pretest scores as a predictor of intercept, or slopes or both. Also should I covary individuals' scores, school scores, or both. I have also been thinking about using a different measure, one that is highly correlated with pretest scores instead of pretest because I dont want to effectively "lose a time point." Is this a viable alternative, and if so do I, again, covary it for individuals, schools, both, and use it as a predictor of intercepts, slopes, or both? I have seen different things done in the literature and also very little written specifically about this. Can anyone help. Thanks much.
Dr. Muthen- thank you very much for the reply. I am not familiar with Multilevelnet, and was not successful when attempitng to search for it either on the statmodel website or on the web generally...could you perhaps direct me to it? Thanks.