Structural equation models were fit to data using Mx (Neale, 1999), a statistical software package designed for structural equation modeling of family-level data. Mx provides an option for use of raw data files as input to accommodate variable missing data patterns and to produce unbiased estimates of the variances and covariances. Model fit was evaluated using a chi-square difference test (i.e., the log-likelihood of a constrained model is subtracted from that of a saturated model with parameters for all the covariance elements and means). Because the chi-square is sensitive to sample size, we used χ2/df < 2 and root-mean-square error of approximation (RMSEA) < .06 as indicators of good fit (Hu & Bentler, 1998). RMSEA takes into account the degrees of freedom of the models, which compensates for the effect of complexity in our multivariate genetic models. The significance of individual model parameters was evaluated using 95% confidence intervals.
