Another limitation is the fact that, for ease of interpretation, only linear models of the data were ultimately chosen for analysis. Our model showed excellent fit to the data, but only when some residual covariances were allowed to be correlated with each other, thus violating an assumption of a true unconditional growth model. The results presented in this model should, then, be viewed as suggestive rather than conclusive. Additionally, although we discuss variables as appearing to mediate the effects of treatment, in the sense that they appear to be links in a causal chain from treatment to outcome, we make no claims for having determined formal statistical mediation.
