Fig. 4.

Parental education has widespread and unique associations with children’s resting state connectomes. We used whole-brain-level (top row), network-level (middle row), and connection-level (bottom row) methods to identify unique effects on children’s resting state connectomes of three SER variables: parental education (left column), household income-to-needs (middle column), and neighborhood disadvantage (right column), Across the three-levels of analysis, parental education (controlling for household income-to-needs and neighborhood disadvantage) demonstrated consistent strong effects. Household income-to-needs (controlling for parental education and neighborhood disadvantage) showed modest, statistically significant effects in multivariate predictive modeling analysis, but showed no significant cells in network analysis and did not deviate from the null hypothesis line in quantile-quantile analysis. Neighborhood disadvantage (controlling for parental education and household income-to-needs) did not show any statistically significant effects in all three analyses. r_cv = cross-validated out-of-sample correlation between actual scores and predicted scores using resting state connectivity data; * = observed correlation higher than all 10,000 correlations in the permutation distribution.