Fig. 4: Causal mediation analysis (CMA).

Mediation models (A, C) and unscaled heatmaps with hierarchical clustering (B, D) for average direct (ADE) and causal mediation (ACME) effects of cortical volume (CV) and thickness (CT) on the relationships between family income (FI) and 36 demographic, socioeconomic, and health behavior variables (X; A, B) and for those of X on the relationships between FI and the morphometrics (C, D) for Discovery (Dis) and Validation (Val) samples of 3892 children each. Separate CMA for selected subsamples assessing ADE and ACME of access to alcohol (N = 3405) and cigarettes (N = 1238), extracurricular sports (N = 2,342), arts (N = 1,927), reading (N = 2,261), and music listening (N = 2229), as well as pubertal estradiol, HSE (N = 1,177), testosterone, ERT (N = 2707) and dehydroepiandrosterone, DHEA (N = 2,811) hormones (X) on the relationships between FI and morphometrics, as well as those of CV and CT on the relationships between FI and X (E); the reproducibility of these pathways was not tested given the reduced size of these subsamples. Schematics highlighting reproducible ADE and ACME for CV (F) and CT (G). ADI area deprivation index, SIB siblings, PED parental education, BMI body mass index, RLE risk of lead exposure, SMA screen media activity. Highlighted variable labels correspond to residential history-derived scores.