An initial, unconstrained model was fit in which substance locations and discriminations were estimated independently for each age-by-sex-by-twin group. Additional models were calculated in which estimates of substance locations and/or discriminations were constrained to be equal across ages and/or sexes. Each model was then also evaluated for whether parameter estimates could be constrained equal across twins. Models were evaluated for goodness-of-fit by examining their comparative fit index (CFI), Tucker-Lewis index (TLI), and root mean square error of approximation (RMSEA). The best-fitting model was identified as the one with the highest CFI and TLI and the lowest RMSEA. (Note: Due to the WLSMV estimator, model fits could not be compared via conventional chi-square difference tests, Muthén & Muthén, 1998–2007.)
