Figure 3. Adjusted cerebellum-to-cerebral-white-matter associations across three cohorts. (A–C) Partial-regression plots for HCP-YA (N = 1,101), DLBS (N = 418), and AOMIC ID1000 (N = 926). Both the standardized focal balance and cognitive score were residualized against the cohort-specific age covariates, sex, log total analyzed tissue volume, and the three remaining orthogonal composition coordinates. Residuals were not restandardized. Every participant is shown; larger points are means in 18 equal-count bins, and lines show the adjusted linear associations. Their slopes equal the reported standardized regression coefficients. The panels have identical axis limits and contain all observations. (D) Cohort-specific coefficients; DLBS and AOMIC intervals use HC3 heteroskedasticity-robust standard errors and residual-df t critical values. The full HCP estimate is descriptive and shown without an interval because family-adjusted uncertainty is unavailable. No pooled regression or HCP independence-based confidence band is shown. The focal balance is ln(total cerebellum/cerebral white matter)/√2.
