The full model (Model I), in which all parameter estimates (a2, c2, and e2) for AN, MDD, and SA, as well as the parameter correlations (ra, rc, and re) for each pair of disorders, was applied to the data. We then applied five nested models: Model II, all common environmental correlations were set to zero; Model III, all common environmental variance parameters and all common environmental correlations were set to zero; Model IV, the common environmental variance parameters and correlations were set to zero and the genetic correlations were set to zero; Model V, the common environmental variance parameters and correlations and the unique environmental correlations were set to zero; and Model VI, the common environmental variance parameters and correlations were set to zero and the genetic variance parameters and correlations were set to zero. The fit of each nested model was compared with the full model. If the difference in χ2 values between the nested model and the full model (Neale & Cardon, 1992), where degrees of freedom (df) is equal to the difference between the df of the
