AFNI 3dNetCorr (Taylor and Saad, 2013) was used to obtain the Fisher Z-transformed 300×300 correlation matrix for each participant. Within-network connectivity was calculated by averaging the correlations of all seeds affiliated with the same networks. Between-network connectivity was calculated by averaging the correlations of all seed-to-seed connections between different networks. From the 8 a priori chosen networks, 8 within-network variables and 28 between-network variables were the outcome variables of the mixed-effects models testing the association between binge/extreme bingeing and change in network connectivity. For each model, the fixed effects of interest were longitudinal standard/extreme bingeing (log-transformed due to skewed distribution). We added additional fixed effects to control for baseline grouping (Control, sBinge, eBinge), baseline age, session (time 1 vs. time 2), baseline age × session interaction, sex, scanner, SES, and frame-wise displacement (FD). Participant identifiers were added as random intercepts. Since the log-transformed longitudinal standard and extreme bingeing were highly correlated (r=.80), these two variables were not entered in the same model. FDR correction for multiple comparison was run separately for each fixed effect of interest.
