I have a question regarding obtaining a negative variance in my growth factor.
I am studying the change of kids' diabetes adherence overtime. When i fitted a univariate growth model for the repeated measures, the model fit was fine and there is no negative variance for the growth factor. However, once i include other predictors in my model, i obtained negative variance for the growth factor. I wonder if this could be due to the fact that the is lack of variability in the diabetes adherence growth factor?
If the variance estimate changes greatly between the run without vs with the covariate, the model may be mis-specified. For instance, perhaps the covariate influences some of the repeated outcomes directly.
Otherwise, as you say, perhaps that growth factor need not be random but can be fixed.