Fig 2. Schematic representation of the hypothesized mediation model.

Treatment allocation is a binary variable representing the treatment (coded 1) and control (coded 0) groups. Loadings for the intercept factor were fixed to 1; the first loading for the slope factor was fixed to zero, and the following to 1.2 and 3.6 respecting the laps of time between the observations. The estimated parameter a_i and a_s represent the effect of the treatment on the initial point (intercept, i.e., 6 months) and on the linear rate of change (i.e., increasing or decreasing) over time of the modeled mediator. The parameters b_i and b_s represent the effect of the mediator (both intercept and growth factors) on the child outcomes. The estimated parameter ‘c’ represents the direct association, i.e., the remaining non-mediated effect. Estimated residual variances for the observed scores are omitted from the figure for clarity.