Prior to analyses, all continuous variables were first centered (average score subtracted from each raw score) to reduce confounds due to multicollinearity, and the binary variable of family history was set to values of 0 and 1. All models were run in Mplus v5.1 (Muthen & Muthen, 1998–2007) with full information maximum likelihood (FIML) estimation under the assumption of data missing at random (MAR) with robust standard errors. In order to increase confidence in the final maximum likelihood values, automatically generated starting values with random perturbations (100 random sets of starting values with 10 full optimizations) were used for all models described below.
