2.3.1. Profile estimation

Models were estimated using robust maximum likelihood estimation and accounted for the multisite design using TYPE = COMPLEX option and weighted according to probability sampling weights. LPA models were estimated on 18 indicators of subcortical volume residuals and OFC thickness measures using Mplus 8.4 (Muthén and Muthén, 1998). Subcortical volume measures were residualized using ICV and standardized to have a mean of 0 and standard deviation of 1 in the LPA. Cortical thickness values were standardized, but not residualized with ICV, following field recommendations (Barnes et al., 2010). Empirical comparisons of models were based on the Akaike Information Criteria (AIC), corrected AIC (AICC), Bayesian Information Criteria (BIC), sample-size adjusted BIC (aBIC), and entropy. Lower BIC and AIC values indicate better fit. Higher entropy values indicate better precision of profile membership assignment. Simulation work (Nylund et al., 2007) found that the BIC performed best of the information criteria. Thus, this criterion is weighted most strongly in empirical comparisons within model sets. We also examined the Lo-Mendell-Rubin Likelihood Ratio Test (LMR-LRT) that compares whether the k-profile solution is a significantly better fit to the data than the k – 1 profile solution. All models were estimated with a sufficient number of random starts to yield a replicated log-likelihood value. All subcortical ROI volume residuals and OFC ROI thickness measures were standardized using the full sample such that interpretations of profiles can be described as deviations in standard deviation units.