Finally, we must make note of limitations to the data analysis of this study. With the availability of only four indicators, the current analyses offered a relatively weaker test of the latent structure of the internalizing disorders. Specifically, the measurement model was over-identified by only two degrees of freedom, and without additional indicators of a possible, second factor (e.g., Hettema et al., 2006), viable alternative models could not be pursued. Further, because of the specifics of the biometric moderation model that we utilized in this study, we did not make use of opposite-sex twin pairs or twin pairs where one of the two twins was missing data on the moderator variable. Hopefully, future improvements in this model will eliminate these reductions in sample size.
