The potential effects of SEX in the model in Figure 2 were further evaluated through a visual comparison of model structures for boys and girls and via invariance procedures. The non-constrained modeling of boys and girls yielded good fit characteristics with CFI = .97, NNFI = .95, RMSEA = .057 (90% confidence interval .40 = .074), and SRMR = .033. For both sexes the paths in Figure 2 for LR to PEER, AGE to PEER, and PEER to ALCOUT were essentially the same, and the path from LR to ALCOUT was similar but ranged from .19 in boys to .29 in girls. As a part of the testing of invariance, constraining relationships as equal for boys and girls revealed that the models for boys and girls were quite similar for path estimates (Π2 (4) = 2.28, p = .69) and variance of the exogenous variable (Π2 (1) = 4.69, p = .10). However, regarding formal invariance testing across sex, there were significant differences for factor loadings on the latent variable (Π2 (2) = 11.71, p<.01). The model fit indices were
