parents were assumed to influence their children rather than vice-versa and, among children, the prospective influences of T1 variables on T2 were also examined. For each SEM model, we assessed model fit using the Chi-square and Root Mean Square Error of Approximation (RMSEA) goodness of fit statistics (MacCallum, Browne, & Sugawara, 1996). We used probit link function and reported standardized coefficients in all the models. The unit of analysis when fitting SEM for the data was the family, rather than individual subjects as in standard regression, to account for the non-independence of data within families. All SEM analyses were performed using Mplus (Muthen & Muthen, 2006). The contribution of variables in statistical models were tested using two-sided tests at the p<.05 level.
