To determine if separate path models were needed for girls and boys, multiple group analysis was used. The full sample model was constrained by gender to test whether the parameter estimates were the same for girls and boys. Parameter estimates varied by gender, as indicated by significant chi-square tests for the structural means, covariances, and residuals (Δχ 2 (10) = 19.23, p < .05; Δχ2(12) = 31.64, p < .01; Δχ 2 (27) = 107.09, p < .001, respectively), suggesting girls and boys should be examined separately.
